What has changed, and what has not.
People now put questions to AI assistants and read AI-written summaries above their search results. Those answers are assembled from pages the system can find, read and rely on. That is new packaging on a familiar requirement: be the clearest, most credible source on the things you actually know.
Google’s own guidance points the same way. It sets no extra requirements for appearing in its AI features beyond those for ordinary search: a page needs to be indexed and eligible to be shown with a snippet. There is no separate switch to turn on.
Help people understand your expertise.
Explain what you do, who it is for and what makes the information credible. Use clear page structure, straightforward language, relevant context and evidence for factual claims.
In practice, that means answering the questions customers really ask, in the words they use, on a page that belongs to that topic. Say who stands behind the content. Give specifics where you can: what is included, where you work, how the process runs, what you do not do. A vague page gives an AI system nothing precise to draw on, and gives a person no reason to choose you.
Keep the website accessible to discovery.
Useful content still has to be reachable. Check the basics before anything more exotic:
- Crawl access. Review robots.txt and any firewall or bot-protection rules. Sites sometimes block search or AI crawlers without anyone having decided to.
- Indexing. Confirm that important pages are indexed, and that a stray noindex tag or a wrong canonical link is not hiding them.
- Delivered content. Look at what the server actually sends. Not every crawler runs JavaScript, so text that only appears after scripts load may never be read.
- Internal links. Make sure each important page can be reached from others through ordinary links with descriptive text.
Structured data should describe what is genuinely present—not turn marketing claims into manufactured facts. Marking up a rating you do not have, or a review nobody wrote, is a liability, not an optimisation.
Measure what you can defend.
Separate observations from outcomes. An appearance in an AI answer is not the same as a qualified enquiry, and a single prompt is not a stable measure of visibility. The same question can produce a different answer an hour later, for a different person, or in a different product.
If you track AI visibility, agree the method first: a fixed set of questions, the same tools, checked at regular intervals, with the results recorded. Read it alongside measures closer to the business, such as enquiries and how those people say they found you. State the limits in the report itself.
Avoid the certainty trap.
Neither search rankings nor AI citations are under an agency’s direct control. Be wary of guaranteed placement, of “AI optimisation” sold separately from the quality of your website, and of reports that show screenshots without a method.
A credible plan focuses on improving the underlying content and experience, with honest reporting about what changed. That work is unglamorous. It is also the part that keeps its value as the interfaces change.
Further reading: Google’s guidance on AI features and websites ↗.
PivotAI perspective. General guidance; not a client case study or a substitute for specialist advice.