Which kind of agency should you hire to get named in AI answers?

Hire the kind of agency that fixes the stage where you are actually losing, and you can work out which stage that is in an afternoon without buying anything. Ask ChatGPT, Google AI Mode and Perplexity the two or three questions a buyer would ask before choosing a vendor in your category. Then look at what comes back and sort it into one of three outcomes.

If the answer names competitors and your pages are nowhere in the sources listed, nothing of yours was retrieved. That is a search and technical problem, and an SEO agency with AI search experience is the right first call.

If your page appears in the sources and the recommendation still goes to a competitor, the engine read you and did not pick you. More page work will not move that. The lever is what the rest of the web says about you, which is closer to the job a communications or PR firm does.

And if the engine will not name vendors at all for your category, or answers with a generic explainer, nobody is winning that question yet. The work is to build the pages and the category language that make the question answerable, and a content agency with subject knowledge can do most of it.

The reason to diagnose before briefing is that all four kinds of vendor now describe their service the same way, and the words AEO, GEO, AI SEO and LLM SEO are used by all of them. Our page on GEO versus AEO covers the naming. What changes between vendors is which stage they can actually reach.

Which stage actually decides whether an AI engine names you?

Retrieval decides most of it, and the condition of your page decides a small part of it. Chen and Koudas, in a study of 158,099 answers and 468,037 citations submitted on 10 October 2026, compared pages within the same website and replayed the same questions through the engines' own interfaces to see what retrieval returned. Which pages the engine's search brings back for the sub-questions it writes for itself explains about 32% of the variation in what gets cited. Page properties explain 2.8%.

A second channel sits beside retrieval, and it is the one that catches marketing teams out. The same study found that ChatGPT includes brands it already knows in the searches it runs, favours sites it recognises, and names brands with no supporting page at all. In controlled tests, changing only the brand name in front of identical product facts changed how the engine ranked the products. The authors' summary is worth keeping: pages are cited when searched, and brands are also named when remembered.

Read those two findings together and the vendor question answers itself. Being retrievable is bought with technical and search work. Being remembered is bought with presence off your own site over a long period, which is brand work whoever delivers it. The page itself is the smallest of the three terms, and it is the term every vendor in this category sells hardest, because it is the only one they can bill as a deliverable.

One caveat belongs with the figures. That study covers commercial questions in two markets over two 90 day collections, with a replication among small B2B vendors. It is the best identified work in this area so far, and it is one study.

What does an SEO agency do for AI visibility, and where does it stop?

An SEO agency fixes the retrieval stage, which on the evidence above is the largest single term. Being crawlable, indexed, rendered without JavaScript errors and linked internally is what puts a page into the pool an engine can draw from, and a competent technical SEO team does that work every day.

Google states the requirement plainly. On its AI features documentation, last updated 10 December 2025, a page must be "indexed and eligible to be shown in Google Search with a snippet", and then: "There are no additional technical requirements." The same page says there is no special schema.org structured data you need to add and no machine readable file to create. For AI Overviews and AI Mode, classic index health is the gate.

ChatGPT has a separate gate, and the controls are not the ones most teams check. OpenAI's crawler documentation states that sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers", that disallowing GPTBot only signals your content should not train foundation models, and that "Each setting is independent of the others." A robots.txt change takes about 24 hours to apply. An agency that blocks the training crawler and assumes it has protected search visibility has the model backwards, and so does one that blocks the search crawler to stop training.

Where an SEO agency stops is the recommendation. It can get you into the retrieval pool and make the page answer the question. It cannot make an engine prefer your name, and most SEO reporting has no way to show whether it did.

What does a PR or communications agency do that an SEO agency cannot?

A PR or communications agency works on the one input that decides recommendations, which is what independent sources say about you. The evidence for that being a separate lever is unusually direct. Lily Ray checked 100 business software queries in Google AI Overviews at three dates between April and June 2026 and found that where a brand's own best-of listicle was cited, the brand was left out of the recommendation in 224 of 323 cases, which is 69%. The competitors named inside those pages got recommended instead.

What separated the two groups was not page quality. Ray reports that recommended brands generally had more referring domains and more brand mentions in AI answers than the brands that were cited and skipped, with at least one named exception she flags herself. She is explicit that the mechanism is hard to confirm. Read beside the brand memory finding in the study above, two different methods point the same way: the engine can read your page and still hand the recommendation to a name it trusts more.

That is the work a communications firm has always done, and nothing about AI search changes the instrument. What changes is the target. Coverage that a journalist would call a small win, an inclusion in a category roundup or an analyst note, is now a source an engine reads at answer time, which makes placement in the pages that already rank for your buying question more valuable than reach.

Where a PR agency stops is your own site. If your pricing, product and comparison pages do not answer the question in the words a buyer types, the coverage lands and the engine still sends the buyer to someone whose page does.

What does a content agency do, and when is a content agency enough?

A content agency buys you the one page level lever that survives statistical controls, which is the match between your page's language and the question being asked. Discovered Labs, analysing roughly 2 million citation observations across 10,000 cited pages, measured prompt to content alignment at a standardised beta of +0.37, with a 95% confidence interval of +0.33 to +0.41 and a false discovery rate q below 1e-73. One standard deviation more alignment was associated with roughly 30% more citations. Two things belong with that number: predictors had to survive four of nine robustness checks rather than all of them, and Discovered Labs sells services in this category, so the choice of what to measure is not disinterested.

Alignment is why our own scoring model puts 35% of the weight on it, more than any other dimension, as our methodology page sets out. A content agency that writes question shaped headings in buyer vocabulary is working on the heaviest page level signal there is.

A content agency is enough on its own in one situation: your technical foundations are fine, your category already has demand, and the gap is that nobody has written the answer. That is more common in young categories than people expect, and it is the cheapest version of this problem to solve.

A content agency is not enough when volume is the strategy. The same study found the format furniture this category sells does very little on its own: an FAQ section at +0.07, a summary block at +0.05, and no significant independent effect for schema markup or Core Web Vitals once domain level differences are controlled for. Twenty more posts will not close a gap caused by your competitor being better known.

What does an AI visibility specialist add that the other three do not?

An AI visibility specialist adds the measurement, and measurement is the part none of the other three can borrow from their existing toolkit. Search Console does not report AI citations, rank tracking does not apply to a surface with no ranking, and a PR clip report cannot tell you whether an engine changed its answer.

Getting that measurement right is harder than it looks, which is the main argument for a specialist. Schulte, Bleeker and Kaufmann ran four AI search engines daily across four campaign verticals for 45 to 46 days, from 24 January to 20 March 2026, and found that the sets of sources cited from one day to the next overlapped by only 34% to 42%. Their bootstrap analysis puts the floor at seven runs per prompt per day for brand level monitoring and eight when source level coverage matters. Anyone reporting your AI visibility from a single check is reporting noise with a decimal point on it.

The engines also disagree with each other. Benjamin Tannenbaum compared 589 citation observations across ChatGPT, Microsoft Copilot, Google and Perplexity on 15 commercial prompts and found 84.9% of engine pairs shared no cited URL at all for the same prompt. That is 15 prompts on one day by a sole author who sells engine measured visibility, so treat it as an upper bound on how disjoint the engines are, not a precise figure. The direction holds: one number across four engines hides four different outcomes.

What our own programme does, as our pricing page describes it, is the page work and the off site work together, with the movement measured monthly on ChatGPT, Perplexity, Google AI Overviews and Claude. We also publish what the method produced on a cold domain in a category we do not sell to, in our pharmaceutical process skid report.

Which one should you hire first, and can one firm do all of it?

Hire against the stage that is failing, and hire the measurement first if you cannot tell which stage that is. Spending a quarter on content when the engines cannot fetch your pages is the most expensive mistake available here, and it is the one that looks most like progress while it happens.

Kind of vendor	Stage it fixes	What it cannot fix	Who it is for
Technical or SEO agency	Retrieval: crawling, indexing, rendering, internal links, crawler controls	Whether an engine prefers your name once it has read you	Companies whose pages never appear among the cited sources
PR or communications firm	Being known: third party coverage, category roundups, analyst and directory presence	Your own pages, where the buying questions get answered	Companies cited as a source but skipped in the recommendation
Content agency	Alignment: pages that answer the buyer's question in the buyer's words	Retrieval problems, and a gap caused by being unknown	Categories where nobody has written the answer yet
AI visibility specialist	Measurement across engines, diagnosis of which stage is failing, and the page work	Earned media at scale, and the time it takes to become known	Companies that need to know what is wrong before they spend

One firm can hold all four stages, and the honest version of that claim is narrower than it usually sounds. A vendor doing the page work, the technical checks and the measurement is doing one job with three parts. A vendor also promising earned coverage at volume is promising a second business, and you should ask which of the two it actually staffs.

If you want the diagnosis before you brief anyone, our free visibility assessment runs the buying questions in your category on all four engines and reports which stage is costing you the answer. Our page on doing AEO in house versus hiring an agency covers the build or buy decision that follows.

What else do people ask about hiring for AI search visibility?

The five questions below come up once a marketing lead has decided to spend on AI search visibility and has to choose what kind of supplier to brief.