How to influence what AI says about your brand

By the CEO of FOUR  ·  An integrated agency, since 1963

The single most important fact about AI search is that you cannot directly edit what it says about you.

There is no admin panel. There is no submission form. There is no Wikipedia-style "request a correction" button. When ChatGPT, Perplexity, Claude or Google's AI Overviews describe your business to a potential customer, they are synthesising signals drawn from across the internet, weighted by criteria you do not control, into a sentence or two that you will probably never see.

That sounds like a problem you cannot solve. It isn't; it is a problem you cannot solve directly! You can solve it indirectly, by changing what AI systems are looking at when they look at you. Most of the work is upstream of the model itself, which is why most attempts to influence AI outputs fail. They are aimed at the wrong layer.

This article is about the right layer. It is structured around the way we approach this work at FOUR, which we call The FOUR A's Method: Authority, Architecture, Amplification and Analysis. Four pillars, integrated. Neither one of them works on its own. Together, they are what it actually takes to shift what AI says about a brand. This is not the next phase of SEO as many digital agenices promote...

The central misunderstanding

Most of what is currently being sold as "AI optimisation" is a checklist of technical tactics inherited from SEO. Add schema markup. Use clear headings. Write 40-to-60 word paragraphs. Lead with the answer. Configure robots.txt to allow AI crawlers. The list is sensible enough as far as it goes. The problem is that it does not go far enough.

Research analysing more than a million AI citations found that over 80% of them come from earned media. Press coverage, trade publications, expert commentary, third-party sources. Less than a fifth comes from brand-owned websites. Which means an agency that addresses only your website is, structurally, working on less than a fifth of the problem.

AI citation patterns are forming now. The brands cited in the first 90 days of a category becoming AI-mediated maintain over three times the citation frequency a year later.

This is the central misunderstanding: AI visibility is treated as a technical optimisation problem, when it is actually a brand presence problem. The technical work matters. It is necessary. It is not sufficient. The agencies offering checklists are answering the wrong question very competently.

The right question is not "how do I optimise my content for AI?" The right question is "how do I make my brand the kind of thing AI systems naturally cite as authoritative when they form a recommendation?" Those are different questions, and they require different work.

Authority. The earned media discipline

Authority is the first pillar of The FOUR A's Method, and it is first for a reason. If the majority of AI citations come from earned media, then earned media is where the majority of the work has to live.

AI systems do not have opinions about brands. They have inputs. When a model is asked which businesses to recommend in a category, it draws on the sources its training and retrieval processes have surfaced as authoritative for that category. Those sources are, overwhelmingly, the same kinds of places human journalists have always treated as authoritative: trade publications, established news media, industry analysts, expert commentary, recognised research. The AI is not making a fresh judgement. It is inheriting one.

Which means the question "how do I get cited by AI?" reduces, in large part, to the question "how do I get covered by the publications AI is reading?" That is a PR question, not an SEO question and it is answered by the disciplines PR has been refining for the better part of a century. Targeted media relations. Genuine expertise positioned in front of the right journalists. Bylined commentary in the publications your buyers and AI systems both treat as credible. Original research that is interesting enough to be reported on. Speaking platforms, panel positions, analyst engagement.

None of this is new. What is new is the realisation that all of it now also influences what an AI says about your brand to a stranger six months later. Earned coverage has always been an investment in reputation. It is now also an investment in machine reputation, and the two compound.

Architecture. The technical and editorial discipline

If Authority is about the inputs AI systems are reading from elsewhere, Architecture is about what AI systems can extract when they read your own content. This is the pillar that most agencies miscall "GEO" and treat as the whole answer. It is not the whole answer. It is, by our reading of the evidence, somewhere around a quarter of it. But within that quarter, the work is real and the discipline matters.

Structure for extraction, not just for reading

AI systems do not read pages the way humans do. They extract information in chunks: discrete blocks of text that can be lifted out, understood as standalone units and incorporated into a generated answer. Content that flows beautifully as a continuous narrative often performs poorly under chunk extraction, because the meaning of any individual paragraph depends on what came before it. Content that is structured into self-contained sections, each of which makes sense in isolation, performs much better.

In practice this means clear headings, ideally framed as the questions buyers actually ask. Direct answers placed in the opening sentence of each section, before any preamble. Paragraphs that are long enough to make a complete point but short enough to be quoted whole. A consistent hierarchy that makes the structure of an argument legible to a system that has never read the page before.

Make the facts machine-readable

The technical layer underneath the writing matters too. Schema markup gives AI systems structured metadata about who you are, what you do, who works for you, where you operate and what you have published. Semantic HTML helps machines understand which parts of a page are the article and which are the navigation. Crawler access has to be explicitly permitted. Page speed and mobile rendering still affect whether content gets indexed at all.

None of this is glamorous and none of it on its own will move the needle dramatically. But Architecture is the pillar that makes the other three work. Without it, even strong earned media and ecosystem presence can fail to translate into citation, because the systems trying to cite you cannot extract what they need.

Amplification. The ecosystem discipline

Authority creates the high-credibility signals AI looks for. Architecture ensures your owned content can be parsed. Amplification is what places your brand across the wider ecosystem AI systems sample from when forming a recommendation.

AI systems do not consult a single source before answering a question. They aggregate across many: the major publications, yes, but also trade press, industry forums, review sites, podcast transcripts, conference coverage, partner content, professional directories, analyst reports, community discussions, academic citations. Brands that appear consistently across multiple categories of source become, in effect, the consensus answer. Brands that appear in one category and not the others tend to be edged out, even when the one category they appear in is strong.

The mechanism is simple: an AI system asked to recommend businesses in a category is, in effect, looking for converging evidence. Five sources mentioning the same brand in different contexts are far more persuasive than one source mentioning it five times. This is why purely PR-led approaches are insufficient on their own and why purely website-led approaches fail even more obviously. The work is inherently distributed.

In practice, Amplification means a deliberate strategy for which third-party surfaces matter for your category. For some businesses it is industry analyst coverage and trade publications. For others it is professional communities and expert directories. For others still it is podcast appearances and conference presence.

There is no universal list. What there is, is a discipline of systematically asking: where does the AI for our category go looking for evidence, and how present are we across those places?

Analysis. The measurement discipline

None of the metrics on a traditional SEO dashboard fully tell you whether AI systems are mentioning your brand, how often, against which competitors, or in what light. Keyword rankings, organic traffic, domain authority, backlink counts: all useful for what they measure, all silent on the question that now matters most. You can have perfect SEO scores and be invisible in AI answers. You can also have the opposite, and never know.

Analysis is the pillar that makes the other three accountable. It is the discipline of running a defined set of buyer-intent prompts across the major AI platforms on a regular schedule, capturing not just whether your brand appears but how it is described, against which competitors, with what framing. It produces a baseline, and then it produces a trend. Without that, you are optimising blind.

You can have a perfect SEO scorecard and be invisible in AI answers. The metrics that matter now are not the metrics on a traditional dashboard.

This is also where most of the rigour of the discipline lives. AI answers vary from session to session, from platform to platform, from prompt to prompt. A serious measurement practice has to control for that variability: the same prompts, the same platforms, fresh sessions, repeat runs where results are borderline, the same competitor benchmarks every time. Otherwise you are measuring noise.

At FOUR, every engagement begins with a Visibility Audit that establishes this baseline. Ten buyer-intent prompts across ChatGPT, Perplexity, Claude and Google AI Overviews, benchmarked against three competitors of the client's choosing. We then re-run the same set on a defined cadence so that improvements, or the absence of them, become visible. It is the simplest part of the methodology to describe and the hardest to do well.

Why none of this works in isolation

If there is one argument worth taking away from this article, it is that the FOUR pillars are not a checklist. They are interdependent and they fail in characteristic ways when separated.

Authority without Architecture produces earned coverage that AI systems struggle to extract or attribute. Architecture without Authority produces beautifully structured owned content that no AI cites because there is nothing else corroborating it. Amplification without Analysis produces activity without accountability: lots of motion across the ecosystem, no idea whether any of it is working. Analysis without the other three produces a clear picture of a problem you have no integrated way of fixing.

The reason most agencies struggle with this work is structural. PR agencies have one of the four pillars in real depth. SEO agencies have a different one. Content agencies have neither at the level needed. Specialist analytics firms can measure but cannot execute. Influencing what AI says about a brand requires all four, working from the same plan, against the same measurement, ideally under the same roof. Almost no agencies are built that way.

FOUR has been built that way for 63 years, since long before integration was a fashionable word in marketing. The discipline AI visibility now requires is, almost exactly, the discipline we have been refining since 1963. The label is new. The work is not.

Where to begin

If you have read this far, the most useful next step is not to start implementing tactics. It is to find out where you currently stand. Most businesses have never tested how AI systems describe them, which competitors are being recommended in their place, or what sources those recommendations are being drawn from. Without that baseline, any work you do is uncalibrated.

This is what the Visibility Audit is for. We test your brand against ten buyer-intent prompts across the four major AI platforms, benchmark you against three competitors of your choosing and produce a written report and a 30-minute walkthrough conversation. It takes us a week. There is no charge for it and no obligation to do anything with the result. Why? Our CEO is passionate about embracing change and supporting those who recognise the need to innovate; this is how we have client engagements spanning 25 years.

If you would like one for your business, the easiest thing is to send us a note. We will take it from there.

FOUR is an integrated agency based in Norwich and London, founded in 1963. The FOUR A's Method is our framework for AI visibility work, structured around four pillars: Authority, Architecture, Amplification and Analysis.