Being Found in AI Answers: ChatGPT and AI Overviews

How AI answers choose sources

When someone asks an assistant for a web agency in Gothenburg that builds Shopify stores, the assistant usually runs one or more searches, reads the top results it can access, and composes an answer citing the pages it used. The pages that get cited are the ones that state the answer explicitly, in text the crawler received, on a site the search index already trusts. Google's AI Overviews work the same way on top of Google's index. So the foundation is ordinary technical and content SEO; what changes is that specificity and directness are rewarded more, and vague marketing copy less.

Crawlability and rendering

AI crawlers are less patient than Googlebot. Several do not execute JavaScript at all, and those that do give a page little time. If your service pages render their text client-side, the assistant may fetch an empty shell and cite a competitor whose HTML contained the words. Check what a plain fetch of your page returns, and check your robots rules, because some sites block AI crawlers by default without deciding to.

Specific, factual content

Assistants quote sentences, not pages. A sentence that says what you do, for whom, where and at what price is quotable; a paragraph about passion for digital excellence is not. Write every service page so that the first screen answers the questions a buyer would actually ask, and write the answers as statements that stand alone when lifted out of context.

• State the service, the location you serve, the typical customer and the starting price in plain sentences near the top of the page.
• Answer real questions under their own headings, in the words a customer would use, so the question and the answer are adjacent in the HTML.
• Give specifics that can be checked: delivery times, what is included, which platforms and integrations you support.
• Keep one topic per page. A page that covers everything is cited for nothing.
• Date and update content; assistants and search engines both prefer pages that are visibly maintained.

Brand mentions and consistency

The model's prior knowledge of your company is built from everything it was trained on: your site, directories, reviews, press, forum threads. Inconsistency there is what makes an assistant hesitate or describe you wrongly. The remedy is not volume but agreement: the same name, the same description of what you do, the same address and organisation number, everywhere.

Measuring AI referrals

Traffic from assistants is visible in your analytics if you look for it, and it is currently small for most businesses, which is a reason to measure it rather than ignore it: it is the trend, not the level, that tells you whether the work above is paying. Set it up once and review it monthly alongside organic search.

1. In GA4, create a channel group or a filtered report for referrals from chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and similar domains.
2. Watch Search Console for impressions on question-style queries, which is where AI Overviews appear most.
3. Ask the assistants your customers' questions yourself, monthly, and record whether and how you are cited.
4. Add a how did you find us option for AI assistants on your enquiry form; self-reported data catches what referral headers miss.

When you do not need this as a separate project: if your site already renders its text server-side, has one clear page per service, and your business details agree across directories, you are doing the work already, and the rest is patience. If any of those fail, they fail for ordinary search too, and fixing them is what our SEO audit and technical fix does, with the AI crawler and rendering checks included; it is priced on the pricing page. To find out where your site stands before spending anything, book a call.

Frequently asked questions