Why AI search decides marketing agencies outcomes
40% vs 34%
third-party sources out-cite firms' own sites
Across our four-engine study of 1,237 citations (all categories), independent listicles took 40% of citations and brands' own sites 34%. For agencies the deciding third-party sources are directories, review sites and 'best agency' roundups.
5 of 713
cited domains shared across all four engines
An agency cited by ChatGPT is usually invisible to Perplexity or AI Overviews, so a single-engine view misleads. You win each engine's sources separately.
#1 on ChatGPT
Reddit was the most-cited source ChatGPT named
In our 25-query study of what ChatGPT cites, Reddit was the #1 source on every query. Marketing and SEO subreddits are where buyers trade agency recommendations, so a firm absent from them is absent from the reasoning.
37.9%
of AI-cited pages rank in the organic top 10
Ahrefs, 4 million AI Overview citations, March 2026. A directory profile or a niche case study can be cited without your homepage ranking, so third-party proof matters more than raw position.
The buyer prompts that decide marketing agencies
| Prompt family | An example buyer asks | What wins the citation |
|---|---|---|
| CategoryBuyer wants options but the query is broad and crowded | best marketing agency | Hard to win directly; narrow it by entering the niche and service families instead |
| NicheBuyer wants a firm that knows their world | best SEO agency for B2B SaaS | Position around one vertical and prove it with matching case studies |
| ServiceBuyer knows the service they need | agency that does generative engine optimization | Own the service term with a clear service page and third-party coverage |
| ComparisonBuyer is choosing between two firms | agency A vs agency B for ecommerce | An honest comparison page plus strong, recent reviews on both names |
| ProofBuyer wants demonstrated results in their space | agencies with case studies in fintech | Publish real, specific case studies a model can quote a result from |
| ReputationBuyer is checking reviews before reaching out | is [your agency] any good | Verified review volume plus addressing the recurring complaint at its source |
Example prompts are illustrative of each family; run your own category, rivals and personas to build the real set.
Which engines matter for marketing agencies, and why
- Primary
ChatGPTWhere many buyers now ask for a vendor shortlist. It often answers well-known firms from memory and reaches for directories and reviews when it searches live, so both your reputation and your profiles matter. - Primary
PerplexityCites its sources inline and rewards a clear positioning page and third-party proof a buyer can click to verify. - Secondary
Google AI OverviewsAppears on commercial agency queries and rides your normal index, so directory and review presence that helps SEO overlaps here. - Secondary
ClaudeUsed in careful vendor evaluation. A case study with specific, verifiable results gives it more to work with than adjectives about your process.
Who AI reads for marketing agencies answers
clutch.coDirectoryThe dominant agency directory, with verified reviews and service tags. A complete, well-reviewed profile is often the deciding source.
reddit.comCommunityThe #1 source ChatGPT cited across all 25 queries in our Reddit study. Marketing and SEO subreddits are where buyers ask for real agency recommendations.- 'best agency' roundupsListicleTrade-publication and blog roundups for a niche or city. Being listed, accurately, puts you into answers your homepage cannot win alone.
- review sitesReview siteG2, UpCity and similar carry aggregate sentiment and service categorization into answers.
linkedin.comSocial proofFounder and agency presence, plus client posts, give a model corroborating signal that the firm is real and active.- your case-study pagesOwnedYour own pages get cited when a case study states a specific, verifiable result, but they are one voice among many.
These source patterns trace to our four-engine AI Citation Study; re-check them as the category moves.
marketing agencies-specific moves
- Claim and prove your directory profilesComplete your Clutch and review-site profiles with the right service tags and gather verified reviews, because those directories are the pages engines read to recommend a firm.
- Pick a niche narrow enough to ownBe the obvious answer for one vertical or one service. The shortlist for 'best AEO agency for SaaS' is far shorter and less defended than for 'best marketing agency'.
- Publish case studies with real resultsAn agency answer leans on proof, so a case study that states a specific, verifiable result gives a model something to quote that adjectives about your process never will.
- Show your own AI visibility as the proofIf you sell AEO, being named when a buyer asks AI for an agency is the demonstration. Track your own citation share and use it as the case study that closes the sale.
Those sections map where you stand: the agency-selection prompt families, which engines matter, and who AI reads to recommend a firm. Prefer (our product) keeps that map current for your own firm, running your agency-selection prompts on ChatGPT, Gemini, Perplexity, Google AI Overviews and AI Mode, and its agency plan runs the same loop for your clients under your brand. One move matters more for agencies than for almost any other model, and it runs through everything below: pick a niche narrow enough that you can be the obvious answer, because a shorter shortlist is a winnable one. The plays follow, in the order a firm starting from low visibility should run them.
The plays that win agency recommendations#
Claim and prove your directory profiles
Complete your Clutch and review-site profiles with the right service tags and gather verified reviews, so the sources AI reads to recommend a firm actually describe you.
Directories and review sites are the pages engines read to recommend an agency, and in our study third-party sources out-cited firms' own sites (40% to 34% of 1,237 citations). A complete, well-reviewed Clutch profile with the right service tags is often the single deciding source for a category or niche prompt. (Prefer AI Citation Study, checked 2026-07-10)
- Claim your Clutch profile and complete it: services, industries, a clear positioning line, and real project entries.
- Add the specific service tags buyers search for, including AEO and GEO if you offer them.
- Run a review drive with recent clients; verified, recent reviews carry more weight than old ones.
- Repeat on the review sites that come up when you run your category and niche prompts.
Done when: Your directory profiles are complete, correctly tagged, and gathering fresh verified reviews.
Verify it worked: Ask a niche agency prompt in ChatGPT and Perplexity and check whether the directories they cite now describe you accurately.
Common failure mode: A thin, mis-tagged profile. Engines read the directory literally, so the wrong service tags keep you out of the exact answers you should win.
Pick a niche narrow enough to own
Position around one vertical or one service so you are the obvious answer for a specific need, instead of one of hundreds for a broad one.
The shortlist for a narrow query is short and lightly defended. 'Best marketing agency' returns household names; 'best AEO agency for B2B SaaS' returns whoever has proven that exact niche, which can be you. Narrowing is the cheapest way to move from absent to named, because you are competing for a smaller answer.
- Choose one vertical or one service where you have the strongest results.
- Rewrite your positioning line and homepage to say exactly who you are for, in the words buyers use.
- Align your service tags, case studies and directory categories to that niche.
- Run the niche and service prompts monthly to see whether you have become the answer.
Done when: Your positioning, case studies and directory tags all point at one clear niche.
Verify it worked: Ask your niche prompt in a grounded engine and check whether you are now named, not just present somewhere.
Common failure mode: Claiming every service for every industry. A firm that is for everyone is the obvious answer for no specific prompt, so it loses every narrow shortlist.
Publish case studies with real, quotable results
Give engines specific, verifiable results to lift, because agency answers lean on proof, not process language.
An AI answer recommending an agency quotes the reason it should be trusted, and a specific result ('grew a client's cited-answer share from 8% to 31% in a quarter') is far more liftable than adjectives about your methodology. Proof prompts and reputation prompts both read case studies over homepages.
- Write two or three case studies with a named metric, a timeframe, and the specific work you did.
- Lead each with the result in one self-contained sentence a model can lift without the rest of the page.
- Keep every claim honest and checkable; an inflated result that a client would dispute is a liability, not proof.
- Add Article schema with visible dates so the recency and authorship are machine-readable.
Done when: Your niche has two or three honest case studies that each open with a liftable result.
Verify it worked: Ask a proof prompt for your niche and check whether your case study, or its result, gets cited.
Common failure mode: Case studies that describe activity without an outcome. 'We ran a campaign' is not a result a model can quote, so it quotes a competitor who states one.
Own your comparison and alternatives prompts
Publish honest comparison pages so a buyer choosing between you and another firm meets a fair, quotable verdict.
Comparison and alternatives prompts are high-intent, and the pages engines read for them are often written by competitors or affiliates. An honest page you control puts your side into the answer, and conceding where another firm genuinely fits better makes the whole page more citable, not less.
- List your top comparison prompts (your firm vs each main competitor, and 'best [competitor] alternatives').
- Write one honest comparison page per major competitor, with a clear one-line verdict a model can lift.
- Name the client each firm actually suits, rather than claiming you win for everyone.
- Add a short, extractable summary at the top: the claim first, the detail below.
Done when: Each major comparison prompt has an honest page of yours that states a liftable verdict.
Verify it worked: Ask the comparison prompt in a grounded engine and check whether your page, or only a competitor's, is cited.
Common failure mode: A comparison that never concedes a point. Models discount pages that read as pure sales copy, and buyers trust them less too.
Earn placements in the 'best agency' roundups
Get your firm listed, accurately, in the trade-publication and niche roundups engines read for category and niche prompts.
Independent listicles were the single largest citation source in our study (40% of 1,237 citations, more than any own-site), and for agencies those are the 'best [niche] agency' guides on trade blogs and directories. One accurate placement in the right roundup can put you into answers your own site cannot win, which is why this is the highest-share off-page lever and worth the outreach. (Prefer AI Citation Study, checked 2026-07-10)
- List the 'best [your niche] agency' roundups that already get cited when you run your niche and service prompts.
- Note which ones omit you or list you under the wrong niche, and prioritize those.
- Pitch the writer with a specific, checkable reason to include you: a real niche result, not a favor.
- Where a roundup is outdated, offer the corrected entry and a recent case study they can cite.
Done when: Your firm is accurately listed in the top roundups for your main niche and service prompts.
Verify it worked: Ask your niche prompt in ChatGPT and Perplexity and check whether the roundups they cite now include you.
Common failure mode: Generic, mass pitches. Editors ignore them, and a placement won by a favor rather than a real result does not survive the next update.
How this fits your existing SEO and new business#
None of this replaces SEO or referrals. The same case studies and service pages often serve both search and AI, and a firm with strong SEO and a real reputation has a head start. The difference is what wins: SEO earns a ranking for a query, AEO earns a citation inside a recommendation, and the citation rewards proof, clear positioning, and third-party corroboration more than keyword coverage. Because most agency-selection prompts have near-zero Google volume, judge this work by citation share across engines, not by rank. And if you sell these services, your own visibility is the proof that closes the deal.
For the full method behind these plays, from the audit to the off-page work that earns most citations, start with what AEO is, or browse every AEO-by-business-model playbook to compare your model with the others.
A worked example
A large educational-content and podcast footprint is the kind of public presence that gets an agency named in 'how do I do X' answers, not just agency-selection ones.
A founder brand paired with a big content library means the agency name surfaces across marketing answers even when the buyer did not start with it.
A clear single-niche position (SaaS and B2B) is illustrative of how narrowing the category makes a firm the obvious answer for a specific need.
Named brands are public, illustrative examples of the category, not customers or endorsements.
Sources
People also ask
- How do marketing agencies get recommended by AI assistants?
- Which prompts should an agency track for AI visibility?
- Which sources do AI engines cite when recommending an agency?
- How does an agency sell AEO or GEO services?
- How is AEO different from SEO for an agency?