Step 1: let the engines read your pages
An engine cannot cite a page it cannot read. Each vendor publishes its crawler’s name and how to allow it in your robots.txt file.
At OpenAI, OAI-SearchBot is used to surface websites in ChatGPT’s search results: a site that blocks it is not shown in ChatGPT search answers. GPTBot relates to model training; the two settings are independent. At Perplexity, PerplexityBot surfaces and links websites in its results. At Google, a page must be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode.
- Check robots.txtOAI-SearchBot, PerplexityBot and Googlebot must not be blocked on the pages you want cited.
- Check your hostGoogle notes that crawling must also be allowed by your CDN or hosting infrastructure.
- Put text in textImportant content must be available as text, not only inside an image.
Step 2: write passages that stand on their own
An answer engine assembles pieces of pages. A cited passage must therefore make sense without the rest of the page: the question in the heading, the answer in the first sentence, the conditions and figures in the same sentence as the claim they qualify.
In practice: one subheading per real customer question, a direct answer right below it, then the detail. Avoid pronouns pointing back to the previous paragraph (“the latter”, “this”) and name things every time.
- One question, one subheading Worded the way your customers would ask it.
- Answer first Two or three sentences that stand alone, before the detail.
- Explicit definitions Each technical term defined where it appears.
Step 3: bring what other pages lack
If ten pages say the same thing, an engine has no reason to cite yours. What sets you apart: your own data, your real examples, your prices, your terms, your dates, and the sources you cite with a link.
Researchers formalised this discipline as GEO (Generative Engine Optimization) in a paper presented at KDD 2024. On their benchmark, they report that such optimisations can boost a content’s visibility in generative answers by up to 40%, and that their efficacy varies across domains. In other words: there is no universal recipe, only tests to run in your sector.
If ten pages say the same thing, an engine has no reason to cite yours.
Step 4: tell the same story everywhere
Before citing you, an engine must understand who you are. It cross-checks your site, your business profile, professional directories, the press and your social profiles. If your name, activity, area and contact details differ from one source to another, it hesitates.
Structured data helps describe your business, on one condition Google spells out: it must match the visible text on the page. Google also asks you to keep your business profile information up to date.
- One name, one descriptionThe same words on your site, profiles and listings.
- Faithful structured dataOrganization, LocalBusiness or Product, aligned with the visible text.
- Up-to-date listingsOpening hours, address, activity: the same version everywhere.
Step 5: date, sign, maintain
Dated information inspires more trust than ageless information. Show the update date, the author or organisation responsible, and revisit your pages when facts change: prices, offers, regulation.
A page that contradicts reality ends up cited for the wrong reasons: an engine repeats what it read.
Step 6: measure, without fooling yourself
The most direct method is to ask the same questions, your customers’ questions, to the engines that matter to you, regularly, and note whether you are cited, how, and who is cited instead. Answers shift as models are updated: a one-off check is not enough.
On Google’s side, visits from AI Overviews and AI Mode are counted in Search Console’s overall search traffic, under the “Web” search type.
For a first read without an account, AudiScale’s free AI scan asks a model the question and compares its answer with your site. The AI visibility feature measures your citation rate on your plan’s engines, at every audit.
What does not help, according to Google
To appear in its AI features, Google writes that no additional requirement or special optimisation is needed: no AI text file, no special markup, no specific schema.org structured data. Search best practices remain the foundation.
Other engines have not published an equivalent rule. The careful path is to get the fundamentals right before stacking devices no one guarantees.
Key points
- An engine cannot cite a page it cannot read.
- An answer engine assembles pieces of pages.
- Before citing you, an engine must understand who you are.
- Dated information inspires more trust than ageless information.
- To appear in its AI features, Google writes that no additional requirement or special optimisation is needed: no AI text file, no special markup, no specific schema.org structured data.
Sources
Official pages read on this guide’s update date.
