Search & AI Visibility for Advertising Agencies: The 7-Rung Citation Ladder

The agency invisibility paradox

There is a paradox at the heart of NAICS 541810 that principals rarely say out loud in front of their teams. We have built an entire industry around making other companies discoverable. We win pitches by demonstrating our ability to lift a client's brand into the answer, into the feed, into the search result, into the shopping list. And then, when a prospect turns to their preferred assistant and asks for a shortlist of agencies to consider for their next brief, most of our shops are entirely absent from the response. The work we do for clients does not accrue to us.

This is not because our client work is invisible; it is because agency-selection prompts and category-answer prompts are structurally different queries with structurally different signal requirements. When ChatGPT is asked "what are the best noise-cancelling headphones under three hundred dollars," it retrieves category education, comparison content, product reviews, and buyer forums. When the same assistant is asked "which agency should I hire to reposition a mid-market fintech brand," it retrieves practitioner content, case studies with named clients, industry press, executive commentary, and directory data. The two prompt families live in almost entirely separate corners of the retrieval graph. Being excellent on the category-answer side does nothing for the selection side.

Our sector has a second complication most sectors do not share. Agencies have always been reluctant to publish. Case studies stay confidential. Client names live under NDA. The best strategic thinking of the shop is walked into a pitch room and then lost. New-business decks are so heavily gated that even prospects have to sign a mutual before seeing the credentials deck. Every one of these instincts is defensible on client-service grounds. Every one of them is a deliberate choice to be uncitable. We have optimized ourselves out of the retrieval graph.

The third complication is that our own websites, more than in almost any other category, are built by people who prize taste over structure. The homepage is a hero video. The work index is a lazy-loaded grid. The case studies are locked behind expand-to-view interactions. The About page is a mood board. This is beautiful. It is also, technically, unreadable to the crawlers that feed models. We are the category that spends the most money on our own websites and gets the least AI-visibility return on that investment. Every principal reading this has seen it in their own analytics: our beautiful sites do not surface us.

What follows is a peer-to-peer briefing for principals who are ready to accept the paradox and fix it. This is not a client-facing white paper. It is a note between shops. We use "we" and "our clients" deliberately — because if you are running an agency in 2026 and you are not already fixing this, your competitors who are will begin taking accounts from you before you know why. The window on this is quietly closing.

How agency selection actually happens now

Ten years ago, agency selection began with an RFP, a referral, or an award-show. All three signals were legible to anyone in the industry. Every principal knew how to work them. The RFP came in through procurement, the referral came through a former colleague, the award-show got you a call from a curious CMO. The signals were slow, expensive, and durable. Shops that were good at them were also generally shops with strong new-business shops.

The funnel now begins with a prompt. A CMO or VP-Marketing at a mid-market brand, needing an agency for a specific brief, opens ChatGPT or Perplexity and types some version of "which agencies are known for repositioning legacy financial services brands with a challenger posture" or "small creative shops based in the U.S. that have done breakthrough work in beverage." The assistant returns four to seven named agencies, each with a short rationale, and the shortlist is set before the prospect has visited a single agency website. They will visit sites, of course — but only the sites of the shops the assistant already named. Being absent from that first response effectively removes the agency from the buying journey.

What matters is not that the prompt happened. What matters is what the prompt did to the funnel below it. Once a prospect has a shortlist from an AI assistant, they do three things very quickly. They cross-check each name against LinkedIn to find the principals and see whether the leadership team looks credible. They open one directory (usually Clutch, sometimes DesignRush, sometimes a category-specific list) to see whether the agency is listed and reviewed. Then they open the shortlisted agencies' own websites for the shortest, most result-oriented review they will ever give: a thirty-second scan for case study proof. If those three checks land, the agency gets contacted. If any of them fail, the agency is silently removed from consideration and the prospect never tells them.

This is the funnel that now delivers the majority of unsolicited inbound to shops with strong AI-answer presence. The RFP and the referral still exist. They are not what changed. What changed is that the population of prospects who reach out at all has been pre-filtered by a machine using signals almost none of us have deliberately optimized. The RFPs and referrals that do arrive have been increasingly informed by AI-answer research the prospect did before you ever heard from them. Even inbound that looks like it came from an old channel is often the tail of an AI-answer conversation.

The strategic implication is that agency new-business is now a two-layer game. There is the layer we already run — positioning, credentials, chemistry, pitch — and there is a new layer above it that decides whether we get to run the first layer at all. Shops without that upper-layer presence are competing for a smaller and smaller pool of briefs against shops that have quietly claimed the shortlist real estate.

What "the perfect agency" prompt actually returns today

Before we get to the framework, it is worth being concrete about what these prompts actually return, because most agencies have not tested them systematically. When we run the shortlist prompts across the four major assistants on behalf of a client, the responses cluster around a small set of patterns.

The first pattern: assistants strongly prefer to name shops they can characterize with a specific reason. Not "Agency A is a great creative shop" but "Agency A is a specialist in early-stage challenger CPG launches and did the recent work for [named brand]." Every named agency comes with a rationale. Shops without a coherent rationale in the retrieval graph do not appear, because the model has nothing confident to say about them. This is why "full-service, all industries, all outputs" positioning has become actively harmful for agency AI visibility — the model has no crisp reason to name a shop it cannot summarize.

The second pattern: the rationale is almost always tied to at least one of five categories — a named client, a documented practice area, a principal's public credibility, an industry-press mention, or a directory position. Shops named without at least one of those five signals in play are unusual. Most named shops have three or four of them working together. Every principal who wants to be more visible in these prompts needs to invest against all five, in some proportion.

The third pattern: the prompts favor recency. An agency that did excellent work three years ago but has published nothing recent, whose principals have no recent public presence, and whose directory listings are stale, drops out of the shortlist even if it is objectively strong. The retrieval systems reward freshness because prospects reward freshness. This is unfair to shops that were dominant a decade ago. It is also the market we now live in.

The fourth pattern: the prompts differ across assistants in ways that are important to understand. ChatGPT in default mode leans on training data and returns shops with deep, older footprints. Perplexity returns shops with recent, well-cited publications and a strong link to the specific vertical. Google's AI Overviews privilege shops with strong classical SEO on the underlying keyword clusters. Claude tends toward precision and returns fewer names but names them confidently. Gemini blends AI Overview logic with the user's own signals. The shared foundation matters most, but the platform-specific tuning matters at the margin more for agencies than for many categories, because agency-selection prompts vary sharply in intent by assistant.

The final pattern — and the one most principals find uncomfortable — is that the models are surprisingly opinionated. Ask them "which agency should I hire" and they will actually rank. They will use words like "particularly known for," "widely regarded as," and "if you want the challenger option, consider." These rankings are not arbitrary. They are consensus reads of what independent sources say. If your shop is not participating in the conversation those sources are having, you will not be in the consensus.

The 7-Rung Agency Citation Ladder

We built the 7-Rung Agency Citation Ladder to give our new-business teams and our client partners a shared vocabulary for what actually needs to be true for an agency to be named in an AI response. Every rung addresses a distinct failure mode we see repeatedly during audits. Every rung is the responsibility of a different discipline inside the shop — which is why one specialist cannot fix the ladder alone. This is a leadership problem before it is a specialist problem.

The seven rungs, from bottom to top, are Provenance, Positioning, Proof, Perspective, Press, People, and Pull. They are numbered because the sequence matters: no shop we have worked with has successfully skipped a lower rung to invest in a higher one. Perspective without Positioning is intellectual noise. Press without Proof is puffery. People without Provenance is confusion. The ladder holds up when the whole of it holds up.

Rung 1 — Provenance: prove you exist

The first rung sounds insultingly obvious to anyone who has been running a shop for a decade. It is not. When we begin an audit of a mid-sized independent, we almost always find at least one broken piece of the provenance layer. The DBA on the site does not match the legal entity in the trade press. The founding year is inconsistent between LinkedIn and the About page. The Crunchbase profile is stale by three years or missing entirely. Two different addresses appear across directories because the shop moved and the old locations were never cleaned. A rebrand from an older name has left a trail of orphaned references. From a human perspective, these are trivialities. From the models' perspective, they are entity-resolution failures. The system cannot confidently decide whether the shop being discussed in a five-year-old Fast Company piece is the same entity as the shop currently listed on Clutch, and so it defaults to caution and does not name the shop at all.

The work at this rung is genuinely unglamorous. Choose the official name and the trading name and enforce them. Publish the legal entity, incorporation year, and address consistently. Own the Crunchbase entry and update it. Register the Wikidata identifier for the shop even if Wikipedia is not yet in play. Maintain a clean Organization schema on the homepage with sameAs pointers to every canonical profile: LinkedIn, Crunchbase, Clutch, Instagram, Twitter, YouTube, and the founder profiles. Every piece of this is checked, quietly, by the retrieval systems when deciding whether the same-named entity in multiple sources is the same shop. Getting it right removes the ambient confusion that keeps otherwise strong shops out of shortlists.

The commonest form of provenance debt for agencies is the rebrand. Every shop we know has been renamed, merged, spun out, or repositioned at some point. When that happens, most agencies communicate the change internally, on the site, and in a press release, and then move on. The trail on the old name lives on in every article, directory, and press mention that predated the change, and the retrieval system never connects the two. The fix is a deliberate campaign of updating third-party sources to the new name, publishing the entity relationship in structured form, and maintaining a permanent legacy page under the old name that redirects semantically and canonically to the new one. Shops that do this cleanly retain their historical citation footprint through rebrands. Shops that do not effectively reset the clock.

Rung 2 — Positioning: earn the sentence

The second rung is where more agencies fail than at any other. Positioning is the sentence the model uses to describe the shop when a prospect asks. If you cannot write it in one confident line, the model will not either — and if the model cannot, it will typically name a different shop that can be summarized more cleanly. This is the single most important consequence of the AI-search transition for agency selection: full-service, generalist, we-do-it-all positioning is now actively selected against.

A useful test: hand five recent case studies to a colleague who does not work at your shop and ask them to write, in one sentence, what your agency does and for whom. If the five sentences they produce disagree materially, the retrieval system will produce a similar disagreement, and it will not confidently name you. Every principal we know has done a version of this exercise and been slightly horrified by the results. The story is rarely as clear inside the shop as it feels.

The right positioning for the AI-search era is a two-part sentence that names a category and a distinctive angle. "A creative shop for challenger CPG brands that need to look and sound bigger than they are." "A brand consultancy for boards navigating category disruption." "A media agency for direct-to-consumer businesses in the growth-to-scale transition." Each of these gives the retrieval system a crisp handle. Notice what they all decline to say. None of them try to claim every discipline. None of them promise every kind of client. That restraint is the point. Specialists get named; generalists get forgotten.

Once the sentence is written, the enforcement work is meaningful but finite. Every website page, every LinkedIn description, every directory entry, every founder bio, every conference speaker bio, every published article's byline note, every sales asset that reaches the outside world — each of these should convey the same core positioning in slightly different words. Not identical language; that reads as robotic. But the same core sentence. Six months of enforcement changes how the models describe the shop for years. Very few agencies do it.

Rung 3 — Proof: publish the work publicly

The third rung is where the entire industry's habits work against us. Agencies have been trained by decades of client-service culture to protect confidentiality — not to publish, not to name, not to quantify. Every reader of this article has had a conversation with a client's legal team about what the agency can and cannot say about a project after it ships. Most of those conversations end with the agency accepting more restriction than the client actually needed. We are the category that under-shares by default.

The consequence is that a large amount of otherwise strong work never appears in the retrieval graph in citable form. A great campaign runs, the trades cover it briefly, and then the case study lives inside the shop as a PDF that only appears in pitch decks. From the models' perspective, that campaign might as well have never happened. The prospect cannot ask an assistant "which agency made that campaign" and get the shop's name back, because the shop never publicly claimed it in a form the retrieval system can parse.

Proof, as a rung, is the discipline of taking every piece of work you are allowed to name and publishing it in the richest, most citable form possible. That means a dedicated URL per case study, not a grid card that opens a modal. It means the client's real name, not "a global technology company." It means quantified outcomes, not adjectives — percent lift, revenue moved, awards won, press earned. It means structured data on the page: CaseStudy schema where appropriate, CreativeWork schema for the campaign assets, Organization schema for the client, and clear author attribution to the strategist or creative lead who ran it. Every one of these signals is a hook the retrieval system can hang a citation on.

Depth over breadth is the correct trade. Ten public case studies, each rich with detail, named clients, and quantified outcomes, outperform forty thin project cards for retrieval purposes. Every rich case study becomes a piece of the shop's identity signal. Every thin card is invisible. The math is worth thinking about carefully before the next annual portfolio refresh: is the shop building forty case-study-shaped things that will not be cited, or is it building ten dossiers that will? The choice is a strategic one, not a design one.

For accounts under NDA, the workaround is real but narrow. Some clients will approve an anonymized case study with the industry, scale, and outcomes intact. Others will approve naming after a delay of a year or two. A small number will approve nothing. In every case, the shop should be aggressive in asking for what it can publish, timed carefully to the client's own communications calendar. The default of "we cannot say" is worth challenging on almost every engagement. In our experience, once the client understands that agency visibility indirectly benefits their own recruiting and vendor selection, they usually approve more than the account team initially assumes.

Rung 4 — Perspective: publish the thinking

The fourth rung is where shops distinguish themselves from executors. Proof shows what the shop did. Perspective shows how the shop thinks. When a prospect asks an assistant for a shop with a distinctive point of view on a category or a discipline, the model is retrieving from a very specific corner of the graph: long-form editorial, published research, opinionated frameworks, manifestos, conference talks that have been transcribed, and podcast appearances that have been indexed. Shops that populate that corner get named. Shops that do not are read as executors, and executors are increasingly commoditized.

The kind of perspective content that moves this rung is not the branded blog post most agencies still publish. Trend round-ups, holiday-timed insight pieces, and "the future of X" articles are effectively invisible to retrieval because the graph is saturated with them and none carry a distinctive angle. What moves this rung is content that says something the industry is not saying yet, or says it with more rigor than anyone has bothered to. Proprietary research that surveys a specific buyer segment. Opinionated frameworks that name the categories and distinctions you use in your own work. Deep post-mortems on campaigns that either soared or underperformed. Long-form arguments about the direction of a discipline. Books, if the shop's leadership is book-shaped.

The quality bar is high. One serious piece of perspective content per quarter, done right, moves the rung meaningfully. Twelve mediocre trend pieces per quarter move it not at all. The shops we see winning are almost always the shops where a principal or chief strategy officer has personally taken responsibility for producing one or two flagship pieces per year that circulate. Delegating this entirely to a content team is a common mistake — content teams produce content, but perspective requires a point of view, and the point of view must come from the people whose taste and judgment the shop actually sells.

Frequency is less important than distinctiveness. A shop that publishes one carefully-researched, opinionated original piece per quarter, and gets it picked up and cited by industry press and other practitioners, will out-cite a shop that publishes weekly puff. Perspective is where the retrieval system learns the shop's shape. Publish shape, not filler.

Rung 5 — Press: earn independent endorsement

The fifth rung is press, and here we mean it in the older sense: coverage by named journalists in publications the retrieval system trusts. This rung is the one most agencies have some existing infrastructure for — PR firms, in-house comms, award submissions — and it is also the rung where most of that existing infrastructure needs to be pointed at different targets than before.

The publications that matter for agency AI visibility are the ones that are heavily indexed, frequently cited by other publications, and structurally readable by retrieval. AdWeek, AdAge, Fast Company, Campaign, Contagious, Little Black Book, Muse by Clio, Ads of the World, Creativity, Shots, Marketing Brew, Digiday, The Drum — these are the tier-one industry sources where a mention accrues to the retrieval graph. Second-tier is the general business press when it covers a specific agency piece of work: Fortune, Bloomberg, WSJ, NYT, The Atlantic. Third-tier, but still worth chasing, is the specialist podcast and newsletter circuit: The Long and Short of It, On Strategy, Uncensored CMO, and the dozens of substack-hosted agency newsletters that are increasingly cited by the models.

Awards are a specialized form of press for our category. The retrieval systems treat the major show databases as trustworthy sources of shop-to-work-to-outcome relationships. Cannes Lions, D&AD, Effie, One Show, Clio, and Kyoorius entries with substantive project descriptions become durable evidence in the retrieval graph. Wins are stronger than nominations, but nominations at the top shows still count. The trade-off for agencies is well-known: award programs cost money, and the industry has periodic waves of skepticism about the returns. The AI-search transition changes the calculus. A well-cased Effie win now feeds not only the immediate PR bump but a durable citation signal that persists in every future selection prompt for the shop's category. That extends the value horizon of award investments substantially.

The mistake to avoid is chasing volume in weak publications. A hundred low-authority mentions do not equal ten strong ones. Focus press investment on the sources the retrieval systems already reference. When in doubt, run the test: search the target publication's domain for a shop you know is well-cited, and see whether recent pieces about that shop show up when you prompt an assistant with a relevant selection query. If the coverage moves the citation, the source is worth pursuing. If it doesn't, redirect the effort.

Rung 6 — People: make the humans visible

The sixth rung is one of the largest and most consequential differences between agency AI visibility and client-side AI visibility. In client categories, the brand is often the primary named entity. In agency categories, the humans are — and the retrieval systems know it. When a prospect asks an assistant to recommend an agency, the assistant often includes principals' names alongside the shop's. "Consider [Shop] — the co-founder [Name] has written extensively on this topic and the shop is known for [work]." That kind of response is only possible when the humans are legible.

Making principals legible is a specific and often uncomfortable set of investments. Every senior person the shop wants credited should have a rich, current LinkedIn presence with a consistent title, a coherent bio, and a public post cadence that establishes them as a voice on the topics the shop wants to be known for. Every founder and executive should have a canonical bio that appears consistently across the site, conference pages, podcast appearances, and press mentions. Every principal who has authored anything of value — a book, a research report, a widely-cited article — should have that authorship clearly connected to their identity through structured data (Author schema) and clean naming. Every principal who has been interviewed should have those interviews indexed and linked back to the shop.

The uncomfortable part for many principals is the LinkedIn cadence. A significant number of senior agency leaders have decided that they do not want to "become a LinkedIn person" — and it is a defensible personal choice. It is also, increasingly, a shop-level cost. The retrieval systems weight active, recent public commentary from principals heavily. A shop whose leaders publish monthly on the topics they want the shop known for will accumulate citation surface much faster than a shop whose leaders never publish. There is no way to fully substitute for principal presence. Marketing and PR teams can support it, editors can polish it, ghostwriters can draft it, but the human being who owns the account has to be recognizable as the source.

The second uncomfortable question is whether the shop should tolerate senior people leaving with their personal brands intact. Historically, agencies have quietly hoped their creatives and strategists would not develop such strong personal brands that they could take clients or attention with them when they left. That posture is increasingly a losing one. In an AI-search world, a shop full of quiet executors is invisible. A shop full of named voices is cited. The right posture is to invest in your senior people's public profiles enthusiastically, and to build a shop culture strong enough that the accumulated equity of their presence stays partly with the shop when they eventually move on. The shops that get this right build succession into the strategy. The shops that get it wrong end up with all their retrieval equity walking out the door periodically.

Rung 7 — Pull: get referenced without asking

The seventh rung is what everything else is really building toward. When clients, alumni, partners, and adjacent voices in the industry mention the shop without being prompted, in venues the retrieval system reads, the citation footprint compounds. This is the rung where the flywheel turns. Every mention makes the next mention more likely, every citation makes the next citation more confident, and every year of that motion widens the moat for future work. It is also the slowest rung to build and the most fragile if the earlier rungs are neglected.

Pull is generated in a small number of specific patterns. Alumni who move on to client-side roles and remain public advocates of the shop where they trained. Clients who publish alongside the shop on shared work — a joint case study, a co-signed research report, a joint conference appearance. Partners in adjacent categories (technology platforms, media companies, production partners) that reference the shop in their own communications. Peers in the wider agency and marketing community who cite the shop's frameworks or thinking. Prospects who mention the shop in comparison content ("we evaluated Agency A, Agency B, and Shop") on their own blogs or in their own podcast conversations.

Every one of these mentions has a common upstream cause: the shop has been generous with the community, has been consistent in its point of view, and has been a good actor across many relationships. Pull is the rung most tied to the shop's culture, and it is the hardest to fake. Shops that have been extractive with clients, closed with peers, or opportunistic with partners rarely build durable pull. Shops that have been open, generous, and consistent build it almost inevitably.

The practical investment in pull looks like: an alumni program that keeps former staff engaged and empowered to talk about the shop; a client-partnership program that produces joint content and joint speaking; a peer network of adjacent shops that you exchange thinking and mentions with rather than compete with; and a public commitment to shared industry infrastructure — conferences, associations, working groups, mentorship programs. None of these are new ideas. What is new is that the retrieval systems now translate them into citation. The shop that has been building this quietly for a decade is the shop the models name.

Credibility signals that specifically move agency citations

Every principal we work with wants a clean answer to the question of which specific credibility signals move the AI-answer needle most. The honest answer is that the signals interact, and no single one is sufficient. But there is a clear hierarchy that emerges consistently across the audits we run. It is worth naming, because most agencies are still spending disproportionately on the lower-weight signals and under-investing in the higher-weight ones.

Named client with quantified outcome. The single most powerful signal for agency retrieval. A case study on the shop's own site that names the client, states what was made, and quantifies what happened is worth more than a dozen mood-board portfolio pages. When we look at the shops most frequently returned by ChatGPT and Perplexity for selection prompts, they almost universally have public, richly-documented case work.

Tier-1 industry press. A single feature in AdAge, Campaign, or Fast Company on a specific piece of work often carries more citation weight than months of the shop's own publishing. This is because the retrieval systems disproportionately trust independent, editorial sources with strong link and citation profiles. A short, generic writeup is worth less than a substantive piece on a specific project.

Major award wins. Cannes, D&AD, Effie, One Show, Clio, and their equivalents are all cited by industry press, listed in structured databases, and picked up by the retrieval systems as durable evidence. Volume of shows matters less than tier of shows. One Cannes Grand Prix does more than ten local award wins.

Directory listings. Clutch, DesignRush, GoodFirms, Sortlist, and their peers are heavily indexed and frequently referenced by AI assistants when producing shortlists. A well-maintained, reviewed listing on the top two or three directories relevant to the shop's category is a meaningful positive signal. A pay-to-play listing on a fifth-tier directory is worth almost nothing.

Principal LinkedIn presence. Consistent, thoughtful publishing by named principals with clear connection to the shop is a strong medium-weight signal. It works best when the principal's posts reference specific work, cite specific evidence, and engage with the wider industry conversation. It works worst when it is diluted with generic motivational content.

Podcast appearances. When transcribed and indexed, well-cited podcast appearances become durable pieces of the retrieval graph. Not every podcast counts equally. The ones that matter are the ones with strong transcripts, credentialed hosts, and audiences that overlap with agency buyers.

Craft portfolio sites. Behance, Dribbble, and Instagram carry surprisingly little weight for agency AI selection, given how much creative time we invest in them. They matter for talent recruitment and for craft credibility with a narrow audience of other creatives. They do not, in our audits, move selection prompts for CMO-tier buyers meaningfully. This is a difficult truth for craft-first shops that have historically over-invested in these platforms.

Content strategy for agencies: the case-study-plus-POV playbook

The content strategy that works for agencies is different from the content strategy that works for our clients, and both differ from the generic advice most content marketing writing offers. What follows is the specific playbook we recommend and use ourselves.

Case studies as the anchor. Every shop that is winning the AI-answer layer has built a well-structured case study library on its own site. Each case study is a dedicated URL, not a modal. Each has a rich hero, a clear brief-and-context section, a substantial work section with images and video, an outcome section with quantified metrics, an attribution section naming the internal team and the client-side collaborators, and a related-work section that links to adjacent case studies. Each is instrumented with schema. Each has been aggressively promoted at launch through the shop's own channels and through third-party press outreach.

POV pieces as the second anchor. Alongside case studies, every winning shop is publishing a small number of substantial, opinionated pieces on the disciplines and categories they want to be known for. These are not blog posts. They are essays, research reports, or manifestos, often between two and six thousand words, with a clear argument, original evidence where possible, and a named author whose credentials are relevant. They are published on a stable URL, preserved indefinitely, and referenced repeatedly in the shop's own thinking and by the shop's principals in their public commentary.

Insight cadence. Below the flagship POV pieces, the shop should maintain a regular cadence of shorter insight pieces — a few hundred to a thousand words — that respond to industry events, take positions on emerging trends, or expand on the flagship arguments. These are less strategically important individually but collectively they signal that the shop is an active voice, and they accumulate into a durable footprint. Weekly or biweekly cadence is a good target for mid-sized shops.

Book, if the shop is book-shaped. Not every shop needs a book. But when a book is possible — when a principal has developed a body of thinking substantial enough to structure into a serious book — the retrieval-graph benefits are outsized. A published book by a shop's founder becomes a durable, deeply-indexed piece of the graph that gets cited by other authors, referenced by press, and used by the retrieval systems as a foundational source. Books are not for every shop. For the shops they suit, they are one of the highest-return investments principals can make.

Newsletters. An earnestly-run newsletter with a real editorial voice, published under a principal's name, is a surprisingly strong contributor to the pull rung. Substack, Beehiiv, or self-hosted — the platform matters less than the consistency and the voice. Newsletters that develop audiences of other practitioners and industry-adjacent voices generate disproportionate citation surface.

What to stop publishing. Generic industry trend round-ups. "Five things we learned at Cannes" pieces without a distinct angle. Holiday-timed puff. Case studies without named clients or quantified outcomes. Reactive content that responds to competitors rather than developing the shop's own argument. Every hour spent on these is an hour not spent on content that would actually build citation surface.

Distribution: creators, industry press, and directories

Content on the shop's own site is table stakes. Distribution — getting mentioned and referenced elsewhere on the reachable internet — is where citation moats are actually built. The distribution work for agencies falls into three main streams.

Creator and podcast partnerships. The industry now has a substantial network of independent creators, podcasters, and newsletter authors who are read and listened to by our buyers. This includes marketing operators sharing their thinking on their own newsletters, agency-adjacent podcasters who host interview shows, YouTube creators covering the industry, and a growing tier of Substack-based analysts. Building relationships with this network — not through PR firms but directly, by being generous with their work and available for their conversations — produces a steady stream of citations that the retrieval systems value highly. Every principal we know who has invested seriously in this network reports meaningful new-business impact within a year.

Industry publications. The traditional trade press still matters and still weights heavily in retrieval. The specific tactic that works best is the substantive contributed piece: a POV essay under a principal's byline, published in a top publication, on a topic the shop wants to be known for. This is different from the syndicated column model and different from the news-hook press release model. It is a piece of thinking, credited to a named person, appearing in a place that will be indexed and cited. Most agency comms teams undervalue this format because it does not produce a splashy immediate PR bump. It quietly moves the retrieval graph more than most other press activity.

Directories and marketplaces. The major agency directories — Clutch, DesignRush, GoodFirms, Sortlist, The Manifest, Agency Spotter — are frequently retrieved by AI assistants for shortlist prompts. A serious listing on the top two or three, with active review solicitation from happy clients, is a meaningful and often-under-invested lever. The trick is that the listing must be treated as a real property, not a checkbox: category filters chosen carefully, capabilities listed accurately, reviews solicited systematically, response rate to inquiries maintained. Directories reward attention. Neglected listings work against you.

What not to bother with. Guest posting on low-authority marketing blogs. Syndication schemes that place identical content on many sites without building unique link equity. Pay-to-play "top agencies" lists on sites nobody with buying authority actually reads. Sponsored placements in publications the retrieval systems ignore. Every one of these is a distraction from the work that would compound.

Retrieval and structured data: what agencies specifically need

The technical retrieval layer for an agency site has some quirks specific to how creative agencies typically build their web presence. Most of what applies to any site — fast time-to-first-byte, clean semantic HTML, working robots.txt and sitemap, no crawl-blocking JavaScript for content — applies to agencies too. But there are five specific issues we see over and over in agency-site audits that deserve special attention.

Content behind interactions. A high proportion of agency sites hide case-study content behind expand-to-view buttons, hover states, video autoplay gates, or animation sequences that must complete before the underlying text becomes readable. Every one of these patterns can prevent crawlers from parsing the content. Rebuilding the case studies as clean, dedicated URLs with the content fully present in the initial HTML is a meaningful engineering task. Every shop that skips it is choosing to be uncitable.

Lazy-loaded work grids. Many agency sites use infinite-scroll or lazy-load patterns for the work index page, which can mean that the shop's own portfolio is only partially crawlable. Ensuring the work index has a paginated, indexable structure alongside the design pattern the shop prefers is a small engineering task with an outsized retrieval benefit.

Schema deployment. The specific schemas that matter for agencies are Organization on the homepage (with sameAs to every canonical profile), Article on POV and insight pieces, CaseStudy or CreativeWork on portfolio pages, Person on principal and team member profiles, and FAQPage on any published Q&A. ProfessionalService schema is a reasonable choice for the shop's overall Organization treatment. None of this is hard to deploy; almost no agency sites do it comprehensively.

Author attribution. Case studies and POV pieces should have clear authorship. The retrieval systems weight authored content more heavily than unattributed content, and they use author credentials to characterize the shop. Every piece of substantial published thinking should have an Author schema block that ties the human being to the shop and to their published body of work.

Crawler policy files. Every agency site should publish an ai.txt and llms.txt at the domain root, explicitly welcoming the crawlers that feed the major assistants. There is no strategic case for an agency blocking AI crawlers — our own marketing content is a marketing asset, not a paid product. Yet a surprising number of agency sites either have no policy file at all or have inherited overly-restrictive defaults from their engineering partners. This is a fifteen-minute fix that removes a real friction.

Different agency types face different visibility problems

The 7-rung framework applies across the sector, but the specific weak rungs and biggest risks vary meaningfully by agency archetype. Understanding the archetype you actually are — regardless of what your credentials deck claims — helps you prioritize the right investments first.

Creative shops

Craft-first agencies with a strong point of view about creative work tend to have the strongest Press and People rungs and the weakest Proof and Positioning rungs. The founders are known figures. The awards case is often strong. The specific work, though, is frequently locked behind design-forward but crawler-hostile portfolio patterns, and the shop's positioning often drifts between "we do beautiful work" (generic) and a specific practice area (specific). The highest-return investments are typically to rebuild the case studies as clean, dedicated URLs with named clients and quantified outcomes, and to write the positioning sentence in specific-not-generic terms. Six months of that discipline moves creative shops materially in shortlist prompts.

Media agencies

Media agencies typically have the strongest Proof rung (they can quantify almost anything they do) and the weakest Perspective rung (they publish little that reads as thinking, versus reporting). The retrieval systems consequently see them as capable but interchangeable. The highest-return investments are typically to develop a distinctive point of view on the discipline — performance versus brand, attention versus impression, first-party versus third-party data — and to publish it in venues the industry reads. Media agencies that make this investment out-cite peers by wide margins within a year.

Digital and growth agencies

Digital and growth agencies (paid media, CRO, SEO, lifecycle) tend to have strong technical rungs but weak Perspective and Positioning. The category is crowded with shops that describe themselves in nearly identical terms. The retrieval systems have a hard time distinguishing them. The winning play is a named vertical or discipline — "we are the growth shop for direct-to-consumer subscription businesses transitioning from paid to organic acquisition" — combined with a serious body of published thinking on that vertical. Generic growth-agency positioning is where most of these shops lose visibility.

Brand consultancies and strategy shops

Brand consultancies typically have the strongest Perspective rung, publishing books and long-form thinking regularly, and the weakest Proof and Distribution rungs. Their case studies are often confidential or presented at a level of abstraction that the retrieval systems cannot verify. Their press coverage is often thin. The winning investments are to negotiate more permission from clients to publish the specific work, and to invest in tier-1 industry press for the thinking rather than only the general business press.

Integrated agencies

Full-service integrated shops have the hardest structural position for AI visibility. Their advantage is breadth. The retrieval systems reward specificity. The resolution is usually to publish distinct practice-area narratives with clear specialists heading each: a named brand-strategy practice, a named creative practice, a named data-and-analytics practice, a named media-planning practice, each with its own identifiable leadership, its own case-study library, and its own POV. Positioning the shop as "one team for everything" without also positioning the specific practices ends up describing an entity models cannot summarize. Practice-area architecture is the fix.

Boutique specialist shops

The category-focused specialists — healthcare shops, financial-services shops, luxury shops, food-and-beverage shops, gaming shops — typically have the easiest structural position of all. Their positioning writes itself. Their case studies naturally cluster in a category the models can characterize. Their thought leadership targets a specific buyer. The mistake we see with specialists is complacency: assuming that being a known specialist is enough. It isn't. Specialists still need Proof, Perspective, Press, and People rungs actively built out. But when they do the work, they compound faster than any other type of shop.

Budget and team allocation for an agency's own program

The awkward question every principal asks eventually: what does it actually cost to run this properly, and who inside the shop owns it? The honest answers depend on the shop's size, and the numbers get uncomfortable when you compare them to how little most agencies spend on their own marketing.

For a small independent shop (fewer than twenty-five people), the program can typically be run inside the existing structure by allocating dedicated time from a principal, a senior strategist or planner, and a mid-level marketing coordinator. External investment is minimal — a monthly retainer for PR outreach, occasional design production for POV pieces, and platform costs for schema and monitoring tools. The main cost is principal time, and it is real: expect a principal to spend a meaningful share of their attention on the program for the first six months.

For a mid-sized agency (twenty-five to a hundred people), the program benefits from a dedicated internal marketing lead who owns the program end-to-end, working with the chief strategy officer on POV and with the principals on press and speaking. External investment scales up: PR firm engagement, content production support for case studies and POV pieces, awards submissions, and platform costs for AI-visibility tracking. Budget in the mid-five to low-six-figure monthly range is typical for shops that want to actually move the needle.

For larger agencies (over a hundred people), the program justifies a dedicated marketing team — typically a marketing director, a content lead, a PR lead, and an analyst — with senior contributions from a chief strategy officer or head of marketing at the executive level. The full program at this scale runs meaningfully into six figures monthly, but the payback in new-business terms is typically several times the investment for shops that execute well. The largest holding companies increasingly have this function centralized at group level, though the retrieval graph rewards individual agency identities more than group brands, so the individual shop marketing budgets remain necessary.

The mistake we see most often is under-investing in the analyst role. Shops that run POV, press, and content programs without systematic monthly measurement optimize blind. A modest analyst investment — even a part-time role or a fractional partner — that produces a monthly report on shortlist-prompt presence transforms the program from hope to discipline.

The 90-day rollout: what to actually do in the next quarter

The ninety-day sequence we recommend and use ourselves looks like the following. It is intentionally aggressive on the first three rungs, because those produce the fastest measurable movement, and it seeds the higher rungs because they compound over quarters.

Ninety days is not enough to finish the work. It is enough to move the lower rungs meaningfully, to put the higher rungs into motion, and to prove enough with measurement to build internal support for the fuller twelve-month program that follows.

Category-specific playbooks: creative, media, brand, growth

Beyond the archetype-level guidance above, each of the major agency categories has a specific playbook we tailor for our client shops. Here is a condensed version of each.

The creative shop playbook

Highest-priority investment: rebuild every published case study as a clean, dedicated URL with named client, quantified outcome, credited team, and CaseStudy schema. The creative work is your greatest asset. Locking it behind design-forward but crawler-hostile patterns is the single most expensive mistake in the category. Second priority: sharpen the positioning sentence away from generic craft claims toward a specific practice area — brand identity for challenger CPG, film for luxury, campaign for entertainment, whatever the actual strength is. Third: activate the principals on LinkedIn with substance, not lifestyle content. Fourth: aggressive award-show strategy targeting the shows that get indexed by the retrieval systems.

The media agency playbook

Highest-priority investment: publish a distinctive POV on a specific media-planning topic where the shop has proprietary data or unusual perspective. Media agencies are drowning in undifferentiated "planning + buying" positioning. A crisp argument — attention over reach, first-party over third-party, always-on over campaign-based — gives the retrieval system something to summarize. Second: turn quantified media outcomes into public case-study format with client permission. The category has more data than any other and publishes less of it. Third: contribute to the trade press on measurement and buying discipline. Fourth: senior planner presence on LinkedIn.

The brand consultancy playbook

Highest-priority investment: negotiate permission with clients to publish more of the specific work. Consultancies suffer disproportionately from confidentiality and can move dramatically by getting even a handful of great case studies published in detail. Second: continue the flagship POV cadence that the category is already good at, but stop hiding it behind PDF gates — publish it as web content the retrieval systems can parse. Third: principal press strategy targeting tier-1 industry publications, not just general business press. Fourth: books, if the leadership is book-shaped.

The digital and growth agency playbook

Highest-priority investment: choose and name a vertical. Digital and growth agencies are the most crowded and most undifferentiated category in the sector. A named vertical — DTC subscription, B2B SaaS, financial services, healthcare — combined with a serious body of thinking on that vertical is the fastest route to shortlist appearance. Second: quantified case studies (this category is naturally good at outcome data; use it). Third: G2, Clutch, and category-directory presence taken seriously. Fourth: principal presence on the specific channels the buyer segment uses, which usually means LinkedIn plus one or two vertical-specific communities.

The integrated shop playbook

Highest-priority investment: architect the shop into named practice areas with named leadership, and treat each practice as its own visibility entity. Integrated shops that publish one shop-wide narrative under a single positioning end up too broad for the retrieval systems to summarize. Second: build separate case-study libraries per practice, cross-linked but individually navigable. Third: separate press strategies per practice, targeting the specific trade publications each practice's buyers read. Fourth: keep the group narrative present but as a wrapper, not as the primary handle.

Common mistakes agencies make with their own marketing

The pattern of mistakes is remarkably consistent across the sector. We have made most of them ourselves at various points. Being aware of them saves months of misdirected effort.

Prioritizing the beautiful site over the citable site. Agencies over-invest in visual design of their own web presence and under-invest in the structural elements that would make the site retrievable. Both matter, but the retrieval elements matter more for new-business impact today. A merely-good-looking site with excellent structure will out-cite an award-winning site with terrible structure every time.

Publishing only when a campaign ships. Many shops publish a case study only when a campaign runs, which produces a lumpy content cadence with long gaps. The retrieval systems reward consistency. Regular thinking published between campaigns keeps the shop present in the graph.

Treating client confidentiality as absolute when it isn't. Most agency confidentiality is defaulted to the maximum without a real conversation with the client about what could be published. A meaningful percentage of confidential work could be publicly credited if the shop asked with the right framing at the right time. Not asking is a self-inflicted wound.

Chasing awards for the awards' sake. Awards are strategic assets when they feed into the retrieval graph. They are expensive vanity when they don't. Every award submission should be evaluated for whether it will meaningfully move citation surface, not just for whether it will produce internal celebration.

Neglecting the principal's public presence. Many shops actively avoid asking principals to be more visible because they have decided the principals do not enjoy it. This is a shop-level cost. The shops that win the AI-answer layer have leadership that has accepted public presence as part of the job.

Confusing awards volume with press coverage. A dozen small-award wins do not equal one substantive press feature. Both matter. The shops that spend disproportionately on awards without investing in press coverage under-perform their potential.

Under-investing in the analyst function. Programs run without systematic monthly measurement optimize blind. Even a modest, part-time analyst role that produces a monthly prompt-panel report transforms the program.

Copying competitor moves instead of building original position. The temptation to look at whichever shop is currently winning the discourse and try to move onto their territory is high. It almost always backfires. The retrieval systems have already learned who owns that territory; the imitator ends up looking like a follower. Build original position instead.

Treating this as a marketing project instead of a leadership program. AI visibility for agencies is a leadership program that touches positioning, publishing, press, people, and culture. Handing it entirely to a mid-level marketing person without executive sponsorship produces predictable results: the technical rungs get built and the strategic rungs languish.

Measurement: what to actually track

The measurement approach we recommend is deliberately simple. A single monthly report against a stable prompt panel, produced by the same person or team every month, using the same methodology. Complicated dashboards are impressive to look at and typically do not survive their second quarterly review. Simple, durable measurement survives.

The core panel is thirty prompts. They should be the prompts the shop's actual buyers would use if they were selecting an agency for a brief the shop wants to win. Ten of them should be broad shortlist prompts ("which agencies are known for X"). Ten should be specific-brief prompts ("we need an agency to reposition a challenger Y brand"). Ten should be discipline-mapped prompts ("who are the best in Z discipline"). The prompts should be stable across months so that the measurements are comparable; add new prompts alongside the existing panel rather than replacing them.

Every month, the panel is run across four assistants: ChatGPT (in default mode without browsing, then again with browsing enabled), Perplexity, Claude, and Google AI Overviews. For each prompt-and-assistant combination, log which agencies are named, in what order, with what rationale. Aggregate across the panel to produce three simple numbers: the shop's presence rate (percentage of prompts where the shop is named), the shop's average position (when named), and the shop's descriptive coherence (how consistent the rationale is with the shop's own positioning statement).

Track those three numbers month over month. Track them by assistant separately, and in aggregate. Track the competitive set alongside — which other shops are named across the panel and how often. The graphs that emerge from six months of this discipline are more actionable than any dashboard we have used from a traditional analytics platform. They tell you exactly which prompts you are winning, which you are losing, which competitors are gaining ground, and where the next investment should go.

Beyond the panel, the second layer of measurement is inbound signal. Which new-business conversations reference AI-answer research? How many prospects arrive already knowing the shop's positioning versus arriving cold? How has the qualified-inbound pipeline shifted over quarters as the program matures? These are qualitative measures that need to be captured through disciplined new-business logging, and they are the numbers that ultimately justify the program's budget internally.

Long-term compounding: why this becomes a moat

The single most important argument for taking this seriously now, rather than eighteen months from now, is that agency AI visibility is one of the most durable moats available in our industry. Every rung of the ladder compounds. Every named client becomes a permanent piece of the retrieval graph. Every POV piece continues to be cited for years. Every press mention persists. Every principal appearance builds durable public presence. Every alumni-driven mention reinforces the shop's identity. The advantage accrues and does not decay quickly.

The mirror image is also true: shops that ignore this for the next two years are not standing still. They are losing ground to the shops that are building. The retrieval graph is not neutral about who has invested. It rewards consistency, freshness, and volume. A shop that is invisible today can still catch up in the next year with an aggressive program. In three years, catching up will be much harder. In five, effectively impossible without buying the presence through acquisition or extraordinary press events.

The competitive dynamic is worth staring at. Every major discovery channel that has emerged in the past twenty-five years — organic search, paid search, social organic, paid social, video, podcast — has followed the same pattern for agencies. Early movers who took the channel seriously when it looked speculative captured disproportionate share and defended it durably. Late movers arrived to a saturated surface where the best they could do was maintain parity at higher cost. AI-driven agency selection is at exactly the early-mover stage right now. The shops that decide this quarter to invest will look, three years from now, like the shops that started their content programs in 2013 or their podcast programs in 2018: quietly indispensable, and difficult to displace.

None of this is a claim that AI visibility replaces anything. Great creative still wins pitches. Great client service still keeps accounts. Great strategy still commands premium fees. What AI visibility does is decide which shops get to compete for the pitch in the first place. That is a strategic filter that operates upstream of everything the shop does well, and if the shop is not present in that filter, everything the shop does well matters less.

A note on how we run our own program

Because this is a peer briefing, it is worth being direct about how we run our own program at Sona & Associates. We are a full-service shop that unites technology, design, and marketing under one accountable team, and every rung of the ladder above is one we work on continuously for ourselves as well as for our clients. Our Provenance is maintained by a monthly checklist of directory reconciliation. Our Positioning is one sentence that every principal can repeat verbatim. Our Proof is a case-study library we invest in refreshing quarterly with rich named-client dossiers. Our Perspective is this article and its companions — substantive published thinking under our principals' bylines. Our Press is a working relationship with the industry trades. Our People are visible on the surfaces the industry reads. Our Pull comes from years of being generous with the community and the discipline.

We do this because we would not be credible advising our clients on this program if we did not run it ourselves. We also do it because it works. Our own inbound has shifted meaningfully in the direction of prospects who arrive already knowing our positioning, having seen our thinking, and having heard our principals on podcasts or seen our work referenced in third-party press. That is the pattern this program produces. It is the pattern we recommend our peer shops build for themselves.

Bringing it together

The advertising agency category is at a specific and uncomfortable inflection point. The buyer journey that used to bring us new-business through referrals, RFPs, and awards is now being pre-filtered by machines that read the retrieval graph we ourselves populate. Shops that populate that graph deliberately are being named in the shortlist prompts that produce the majority of qualified inbound today. Shops that are not are being silently omitted from those prompts and never told.

The 7-Rung Agency Citation Ladder — Provenance, Positioning, Proof, Perspective, Press, People, Pull — is our operating framework for fixing this. Each rung addresses a specific failure mode. Each is owned by a discipline the shop already has. The work is unglamorous but not exotic. It requires a sustained principal-level commitment for six to nine months to move the lower rungs and a longer, patient investment to move the upper ones. It rewards shops that stay consistent and punishes shops that treat it as a project rather than a program.

Every principal reading this has a choice about which side of that dynamic to be on. The window is still open. The competitive field is still uncrowded relative to what it will be. The investment required is meaningfully less than what most shops already spend on client-acquisition activities that produce less durable results. The ladder is climbable.

If you want a partner to climb it with you, that is what we do — brand, build, and growth on one accountable team, with AI visibility as the connective tissue between all three. If you would rather run it in-house, this article is the plan. Either way, the shops that start this quarter will be the shops that are recommended by the AI assistants your prospects use two years from now.