When a buyer asks an AI assistant what to use, the answer is a short list of names. Most brands treat getting onto that list as a mystery, or as a trick someone has figured out and is selling. It is neither. An engine recommends a brand when it can describe that brand clearly for the question asked, finds the description confirmed in sources it trusts, and has no contradictions that make it hesitate.
Nothing below guarantees a recommendation, and nobody can honestly promise one. What this work does is remove the reasons an engine has to leave you out.
Engines recommend what they can describe with confidence
An AI engine names the brands it can connect to the question with the least doubt. Clarity on your own site, agreement across other sources and the absence of conflicting facts all feed that confidence.
Engines that search at answer time, as Perplexity does routinely and ChatGPT, Gemini and Claude do when they browse, retrieve pages and compose from them. When an engine answers from training data, it relies on what it absorbed earlier. Most answers mix the two. Either way, the brands that get named are the ones that clearly fit the request and that several sources agree on.
Decide what you want to be known for
Pick one specific thing you want engines to associate with your brand, and write it as a single sentence a buyer would recognize. Every other step in this guide depends on it.
“The platform for modern teams” matches every competitor and no real question. “Scheduling software for independent physical therapy clinics” is something an engine can connect to a prompt.
- Category. Use the words buyers use, not your internal product language.
- Audience. Team size, industry, role or situation, as specific as you can honestly be.
- Reason. Why you are the right choice for that audience, in terms a page on your site can back up.
Then list the prompts where that position should earn you a place: category questions, comparisons with competitors, and the problem descriptions buyers type before they know your category exists. Those prompts are what you will measure against.
How do you make your own pages easy for AI engines to use?
State the facts that matter plainly, near the top of the relevant page, in sentences that still make sense when quoted alone. Engines lift passages, not pages.
Answer before you argue
For each question a buyer asks about your product, there should be a page where the first sentence under a heading answers it. What it does, who it is for, what it costs, how it compares, what it integrates with. An answer buried after three paragraphs of scene-setting often goes unused, and the engine takes its answer from somewhere else.
Mark it up honestly
Structured data tells machines what a page is about without interpretation. Organization schema clarifies who you are, Product or SoftwareApplication schema describes what you sell, and FAQ schema marks up questions and answers that genuinely appear on the page. Markup that does not match the visible content is worse than none.
Say the same thing everywhere
Your homepage, pricing page, documentation and profiles should agree. If one page says you support a feature and another implies you do not, the engine has a reason to hedge, or to trust a third-party description over yours.
Make pages easy to trust
Use a clear heading hierarchy with headings phrased as buyer questions, add author bylines where expertise matters, show genuine update dates, and build comparison and use-case pages for the prompts on your list. An llms.txt file that points AI systems to your most useful pages is a small addition worth making. For the page-level detail, see how to get cited more often.
Build a page for every prompt that matters
If a prompt on your list has no page on your site that answers it, an engine has nothing of yours to use for that question. Map each prompt to a page, and create the page where the map has a gap.
Comparison pages
Buyers ask engines to compare named products all the time. A page that compares you with a named alternative, fairly and with facts a reader can check, gives the engine a direct source for that question. Say where the alternative is the better fit. Engines and buyers both discount comparisons where one side always wins.
Use-case pages
Problem prompts describe a situation rather than a product. A page for each situation you genuinely serve, written in the words a buyer would use to describe it, connects your brand to questions where competitors are often absent because they only built pages for the category term.
Pricing and plan pages
Pricing questions are among the most common prompts about any product, and among the most often answered wrongly. State your pricing model plainly in text, explain what each plan includes, and keep it current. If you do not publish prices, say how pricing works and how to get a quote, so an engine does not fill the gap with a guess from an old article.
Pages that answer objections
Prompts like “is it worth it” or “what are the downsides” get answered whether you publish anything or not. A straightforward page on limitations, security or migration effort gives the engine an accurate account to work from instead of the most negative thread it can find.
Third-party sources carry more weight than your own site
Engines frequently describe and recommend brands based on what other sites say about them. Being present and accurately described in those sources is often the difference between being named and being left out.
From the engine’s side this is reasonable. A brand calling itself the best choice is not evidence. Several independent sources describing it as a strong fit for a specific need is.
Find the sources that matter in your category
Run your prompts through each engine and record the cited URLs. The same review platforms, comparison articles, community threads and documentation sites tend to recur. That list is your outreach plan, and it is rarely the list you would have guessed.
Earn inclusion rather than buying it
- Review sites. Keep your profile complete and current, and ask real customers for reviews. Never buy or fabricate them.
- Comparison articles. Contact authors who compare your competitors and leave you out. Offer accurate information and product access, not payment for placement.
- Communities. Answer questions where your buyers ask them, helpfully and with your affiliation disclosed.
- Documentation and directories. Make sure integration partners, marketplaces and industry directories list you with correct, current details.
This work is slow and does not scale the way publishing does. It is also the part competitors most often skip, which is exactly why it tends to pay.
Correct wrong facts at the source
When an engine gets your brand wrong, fix the page the error came from, not just your own site. An engine repeating an old price or a discontinued feature is usually reflecting a page that still says so.
- Record the wrong claim, the engine and the prompt that produced it.
- Check the sources cited for that answer to find where the claim came from.
- Make sure your own pages state the correct fact clearly and somewhere obvious.
- Ask the publisher of any outdated third-party page for a correction, or update listings you control.
- Re-run the prompt after the sources change, and keep checking.
Errors that come from training data rather than live sources take longer to change. The full process is in what to do when AI gets your brand wrong.
Should you allow AI crawlers on your site?
Allow the AI crawlers you want reading your site, and block only the ones you have a deliberate reason to block. An engine that cannot fetch your pages cannot quote them.
Plenty of sites block AI crawlers by accident: a broad robots.txt rule, a firewall setting, or bot protection that treats every unfamiliar user agent as hostile. Check those settings against the crawlers used by the engines you care about, then check your server logs to confirm the crawlers are reaching the pages you want read. Blocking some crawlers can be the right call, for paid content for instance. Make it a decision rather than a default.
What should you avoid?
Avoid anything that tries to manipulate the engine rather than inform it. Tactics that misrepresent your product or your coverage tend to backfire with engines and buyers alike.
- Fake or incentivized reviews. They break most review platforms’ rules, and a profile that gets cleaned up is worse than a small honest one.
- Hidden instructions aimed at AI. Text that humans cannot see is deceptive, and engines have every reason to discount pages that try it.
- Mass-produced thin pages. Near-identical pages for slight prompt variations add noise without adding facts.
- Claims nothing supports. Calling yourself the category leader gives an engine a statement no independent source confirms.
Why should you measure each engine separately?
ChatGPT, Gemini, Perplexity and Claude use different sources and reach different conclusions from the same question. A blended score hides that you might be strong in one and absent from another.
Run the same fixed prompt set in each engine on a regular schedule, and for every answer record whether you were named, where you appeared, how you were described and which sources were cited. Then compare engines side by side.
- Absent in one engine only? Look at what that engine cites. It may lean on a source where you have no presence.
- Named but described poorly? Check the cited sources for outdated or negative claims, and check how your own pages state the facts in question.
- Named everywhere but low in the list? Work on the comparison content and third-party coverage that would make you the obvious fit for that prompt.
Noma tracks ChatGPT, Gemini, Perplexity, Claude and Grok, asking each prompt once a day from a fresh session and storing the full answers, so the per-engine view and cited sources are there to compare.
How long does it take to get recommended?
Expect months rather than weeks. Some fixes, like removing a crawler block or clarifying a key page, can show up sooner in engines that search the web. Third-party coverage and anything that depends on training data takes longer.
AI answers also vary between runs, so one good or bad day tells you little. Judge progress on the trend across a stable prompt set, and leave the set alone so the comparison holds.
What to do first
In rough order of return on effort:
- Write your positioning sentence and the prompt list that tests it.
- Check that the AI crawlers you want can actually reach your key pages.
- Rewrite the openings of those pages so each one answers its question in the first sentence, and remove contradictions between them.
- Correct the wrong facts engines are already repeating, starting at the source.
- Map the third-party sources cited in your category and work on being accurately included.
- Measure per engine every month, pick the biggest gap, and repeat.
None of this is a shortcut, and that is the point. Engines recommend brands that are clear, consistent and confirmed by others. For a structured walk through the whole discipline, the AEO guide covers each stage in more depth.