AI engines do not have an opinion about your company. They assemble one from what other sources say about you.
That fact reorganizes the problem, and not conveniently. Most of the work that earns a recommendation happens somewhere other than your own website, which is the only part of this you fully control.
What the research actually found
Researcher Amit Prakash Sharma ran 2,240 questions about 112 companies through ChatGPT and Perplexity, then compared which surfaced against what they had done to be found.
The techniques marketed as AI optimization showed no measurable relationship to whether a company appeared. What did predict visibility were two unglamorous things: how many credible outside websites linked to it, and whether it had real presence in the communities where its category gets discussed. His conclusion was blunt. Build the foundation, and visibility follows.
Here is the uncomfortable version. Much of what is sold as a new discipline is authority building with a new label, and the relabeling happens because authority building is slow and hard to sell as a subscription.
This is a public relations problem in an SEO costume
If credible outside links and community presence move the needle, the levers are earned rather than built: trade publications, industry associations and directories, review platforms, local media, and real participation where your category gets discussed.
There is a shortcut to knowing which of those matter for you. When you check how AI engines answer your category’s questions, note which websites they cite. Those citations map where your reputation is being formed, and they are rarely the properties you would have guessed. That is your target list, and one of the few places where measurement hands you the work order directly.
The same logic applies to authorship. A page with a named author, real credentials, and a footprint outside your own domain corroborates more easily than the same words under a logo. That means real bylines in places you do not own, not a grid of headshots on the about page.
Specific beats broad, every time
Discovery questions are specific. People do not ask who is good at marketing. They ask which banks are best for a construction business, or what the strongest online nursing programs are for working adults.
If you have no page unmistakably about that service, for that buyer, in that market, you are not a candidate to be named.
This is the most common structural failure we see. A company offers eleven services across four markets and has one page describing all of it. That page is about nothing in particular, so it competes for nothing in particular.
The fix is work, not mystery. A dedicated page for each service line, each market, and each audience you want, written in the language a buyer would use, not the language of your capabilities deck. And write them as answers. If a prospect asks what to consider before choosing a provider, there should be a page answering that directly, near the top.
Your facts have to agree with each other
When your information conflicts across the internet, engines tend to leave you out rather than describe you wrong. Uncertainty is easier to resolve by omission.
So go find the conflicts. Check locations and hours, service lists, leadership, credentials, and performance claims. Check your website against your directory listings, association profiles, review pages, and the landing pages nobody has opened in three years. Old campaign pages are usually the culprit, abandoned while still ranking.
It is the least interesting work on the list and the highest return per hour. You are removing reasons to skip you.
Why you cannot just buy this
Look at what the list requires. Earned media. Content strategy and writing. Technical website work. Review and reputation management. Web development.
Five functions. A vendor selling an AI optimization subscription is selling one slice of one of them, usually the technical slice, which the research suggests matters least. It fits no single job description or line item, which is why this work sits unowned in most organizations.
A ninety-day sequence
Weeks one and two. Establish a baseline. Run your category’s buying questions across the major engines, using several phrasings rather than repeating one. Record whether you appear, not where, and log every website cited.
Weeks three and four. Fix your facts. Reconcile every conflict across your site, listings, profiles, and abandoned pages.
Weeks five through eight. Build the specific pages. One service, one audience, one market each, written as answers.
Weeks five through twelve, in parallel. Earn five to ten credible third-party mentions on the properties your baseline showed are being cited. Not press releases. Real placements.
Week twelve. Re-measure. Same questions, same phrasings, same method. Compare presence, not position.
Ninety days moves a specific category, not a broad one. You will not become the answer for “best marketing agency” in a quarter.
The through line
AI engines know your company exists, and that is not the same as recommending you. The tools promising to score your position are measuring something that will not hold still. The work that closes the gap is the slow work of being genuinely well regarded by people other than yourself.
There is no shortcut in that, which is inconvenient now and precisely why it holds up later. Anything a competitor could buy next month was never an advantage.
Want to know what AI engines say about your category, and what to do about it?
In a Brand Awareness Consultation for AI Engines, we establish your real baseline across ChatGPT, Gemini, Perplexity, and Google’s AI search, identify which outside sources are shaping the answers, and give you a prioritized plan across content, public relations, reputation, and website work.
Sources
Sharma, Amit Prakash. “The Discovery Gap: How Product Hunt Startups Vanish in LLM Organic Discovery Queries.” arXiv. Based on an M.Tech thesis at the Indian Institute of Technology Patna. https://arxiv.org/abs/2601.00912