AI visibility is whether a company can be found, understood, verified, and trusted by the systems people now use to ask questions — and by the people reading those answers. That's the whole definition. It is not a channel you buy, a ranking you win, or a switch you flip.
What a company controls is narrower than the marketing around it suggests, and larger than most owners assume. You control what you publish, how clearly you describe yourself, whether machines can read your site, whether independent sources corroborate you, and whether any of it is kept current. You do not control what any AI system says on any given day. The useful work lives entirely on the first side of that line.
The vocabulary, cleared up once
Four terms are circulating, three of them are mostly the same thing, and the confusion is doing real commercial damage — mainly by making a measurable discipline sound like a secret.
SEO — search engine optimization. Making a site findable and credible in conventional search. Decades old, well documented, still load-bearing.
AEO — answer engine optimization. Industry coinage. It has no documented origin and no agreed definition; treat anyone claiming to have invented it with the same skepticism you'd apply to any other unfalsifiable credential.
GEO — generative engine optimization. This one has a real academic origin: a paper submitted in November 2023, revised through 2024, and accepted to KDD 2024. Worth knowing what it actually found, because the popular version is distorted. Its headline figure — visibility gains "up to 40%" — is a ceiling, not an average. The per-method averages ran roughly 20–27%. Adding quotations, statistics, and cited sources helped most. Keyword stuffing helped least. And the most interesting result rarely gets quoted: lower-ranked sources gained enormously from these changes while top-ranked ones sometimes lost visibility. The authors framed it as democratizing. It also means the tactics that help a challenger brand may not help a category leader.
The caveat matters as much as the finding: that study tested generative engines as they existed in 2023 and 2024. It is not evidence about how Google's AI Mode or ChatGPT's search behaves today.
AI visibility — the outcome all three are pointed at. We use it because it describes the result rather than the tactic, and because it survives the next rename.
Here is the part the acronym economy would rather you didn't know. Google's published guidance for site owners states: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." It goes further — "You don't need to create new machine readable files, AI text files, or markup to appear in these features," and "There's also no special schema.org structured data that you need to add." Google's own framing is that optimizing for generative AI search is optimizing for the search experience, and thus still SEO.
That is not an argument that nothing has changed. It's an argument that what changed is what good looks like, not the existence of a new secret channel.
The four things it rests on
Found
A system has to be able to reach your pages. That means indexable, crawlable, and not accidentally excluded. This is the cheapest thing on the list to check and the most embarrassing to get wrong.
It's also more nuanced than "allow AI or block it." OpenAI alone publishes three separate crawlers doing three separate jobs — one for training models, one for surfacing sites in ChatGPT's search features, one for user-initiated visits — each controlled independently. Google now offers site owners a Search Console control to opt out of generative AI features specifically, and states that opting out is not used as a ranking signal elsewhere in Search. These are real, meaningful choices. Most companies have never made them consciously.
Understood
A system has to be able to say what you are. Not just index you — describe you. That requires that your name, category, location, and offering read consistently across your own site and everywhere else you appear, and that your product and service pages contain enough substance to summarize.
Entity clarity is unglamorous and it's usually where the actual problem lives. A company with an ambiguous name, three inconsistent category descriptions, and a name it shares with a business in another industry will be described badly no matter how much content it publishes. This is worth checking before anything else, because it is the one failure that makes all the other work invisible.
Verified
Independent corroboration. Press, reviews, trade coverage, credible directories, genuine third-party discussion. Systems weight these for the same reason a careful buyer does — a company describing itself is an interested party.
This is why AI visibility and public relations turn out to be the same project wearing different clothes. Earned coverage isn't only awareness anymore. It's evidence, and evidence is what gets cited.
Trusted
The accumulated signal of a company that is consistent, current, and substantiated. Freshness is a measurable part of this — an analysis by Seer Interactive of roughly 47,000 citations across ChatGPT, Gemini, and Perplexity this spring found about three-quarters of cited pages had been updated within the past year. The same analysis found something more useful for planning: refreshed older pages outperformed genuinely new content. It's vendor research rather than peer-reviewed work, so hold it loosely — but the practical implication is sound. Maintaining what you have beats publishing more of it.
What you control, what you don't
Inside your control: site architecture and crawl access · entity consistency across every surface · depth and accuracy of product and category information · language matched to how buyers ask rather than how you catalog · structured data where it earns its keep · third-party coverage and mentions · original, useful published material · verifiable proof · currency of all of the above.
Outside anyone's control: what a given model outputs for a given user on a given day · whether a citation persists · inclusion in every relevant answer · a ranking position inside a generated response · how any platform changes its systems next quarter.
The honest way to say this: you cannot buy an outcome, but you can materially improve the odds, and you can measure whether you did.
How to make it measurable
We know what the failure mode looks like in this category because we went and tested it.
During development of LRBG’s AI Visibility methodology, we tested seven widely circulated AI-visibility prompt approaches across sixteen blind runs, using one AI system with live web access, independent replication, an evidence log on every run, and a separate skeptical evaluator scoring each output against two live agency websites afterward. The outputs were more careful than we expected on facts: 34 of 35 spot-checked factual claims held up.
Two things were wrong with all of them.
The first was fabricated precision. Every one of the seven demanded numbers the system could not possibly know — priority scores, commercial-value estimates, overall audit scores with no rubric behind them — and the system obligingly produced them. Two runs of the same audit on the same site returned scores of 58 and 62 out of 100 while disagreeing about the single most consequential technical fact on the site. Converging scores concealed diverging evidence, which is worse than an obvious error.
The second was more fundamental. Not one of the seven ever measured AI visibility. They generated recommendations. None of them queried an AI engine to establish whether the brand was actually cited before the work, or after it. A methodology that never takes a baseline cannot tell you whether it worked, which means it also cannot be wrong.
One limitation on our own study, stated plainly because the point of this section is evidence discipline: we ran it on a single AI system with live web access, not across every assistant. It tests the general validity of those approaches, not any one product's behavior.
So our method is deliberately boring:
Establish a baseline. Record how the brand currently appears, before anything changes.
Test real buyer questions. A representative set, in buyers' language, across the tools they use.
Check technical accessibility. Can systems reach the site, and is that a decision or an inheritance?
Evaluate entity understanding. Can a system correctly say who you are and what category you're in?
Evaluate authority and evidence. What independent corroboration exists?
Identify gaps — prioritized against what a fully visible brand in the category would show.
Strengthen the signals within your control. This is the only step that changes anything. Everything before it is measurement.
Retest. Same questions, same tools, same format, on a schedule.
If a proposal you're evaluating has no step 1 and no step 8, it isn't a program. It's an opinion with an invoice attached.
The sane position
AI visibility is neither the end of search nor a new dark art. It is the same discipline that has always decided whether a company gets chosen — clarity, evidence, accessibility, and trust — now being read by a second audience that summarizes rather than lists.
That second audience is worth taking seriously. Pew Research reported this June that 49% of US adults use AI chatbots and 42% use them to search for information. Adobe's retail data put the share of consumers who have used an AI assistant while shopping at 39% as of spring, and found AI-referred visitors converting better than other traffic by March — a reversal from a year earlier, when they converted worse. That reversal is the most useful thing in the data, and not for the reason it's usually quoted: it tells you the ground is still moving, and that anyone presenting a single snapshot as a permanent truth is guessing.
Taking it seriously means measuring it, not mystifying it.
If your brand is absent, the fix is almost never exotic. It's usually clarity you never wrote down, corroboration you never pursued, or a technical default you never chose.
We test AI visibility before we recommend anything, and we retest afterward — so the work has a before and an after rather than a promise. See how we test it. If you want the wider picture of how search, content, PR, and site architecture fit together, that's what we build — and if you'd rather start with a diagnosis than a proposal, start with a Growth Audit.