Why does AI search work differently from Google?
Google crawls your website and ranks pages in near real time. Changes you make today can appear in Google results within hours. AI search tools work differently. They pull answers from a knowledge base compiled from the web and refreshed at intervals. For most AI tools, that refresh cycle sits somewhere between three and six months.
This creates a lag that catches most businesses off guard. If your website content was thin, unsigned, or poorly structured when the last refresh happened, you were not included. No amount of last-week optimisation will get you into an AI answer sourced from three months ago.
There is also a two-phase dynamic. Phase one: AI gives the customer an answer and names brands. Phase two: the customer searches that brand name in Google to verify it. If you are not named in phase one, you never get to phase two. Monitoring brand search volume in Google Search Console is now one of the most important signals you can track. You may also see this framed as AEO (answer engine optimisation) or GEO (generative engine optimisation), but the goal is the same: being cited in AI-generated answers.
What does it actually take to be cited by ChatGPT, Perplexity, or Google AI Overviews?
Six things matter, in order of importance.
1. Content that signals real expertise. AI tools weight named authors, specific examples, and demonstrated track records. An unsigned article that says AI search is changing things will not be cited. A signed piece from a named person, citing specific data and real outcomes, has a much better chance. This is what Google calls E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness.
2. Structured headings that match how people ask questions. AI answer engines scan for headings that match real question patterns and pull the section that follows. H2s phrased as questions perform better.
3. Strong internal linking and topic authority. A site with tightly linked content in a specific subject area signals more topical authority than a site with lots of isolated articles. Identify your core topic and make sure your key pages link to related content via considered internal linking.
4. [Schema markup](/glossary/schema-markup). FAQ schema, Article schema, and Organisation schema reduce friction for AI tools trying to understand your content. They do not guarantee citation, but they remove obstacles.
5. Brand mentions across the web. AI tools look at what the rest of the web says about you: trade publications, industry roundups, podcast show notes, news articles. A single well-sourced article on your domain, mentioned by three or four external sources, is worth more than ten isolated articles with no external footprint. Pair this with NAP consistency across directories.
6. A clean technical foundation, starting with your robots.txt. This one surprises almost every business I speak to.
Is your website accidentally blocking AI from reading it?
The most common technical mistake right now is sites blocking AI reading crawlers, often without realising it. Cloudflare's AI Fight Mode, increasingly switched on by default for new accounts, blocks a wide range of AI crawlers including ClaudeBot (Anthropic's crawler), GPTBot (OpenAI's crawler), Googlebot-AI (Google's AI Overview crawler), and PerplexityBot. The same problem appears in robots.txt.
The key distinction: blocking AI training crawlers (bots that scrape your content to train future AI models) is legitimate and often desirable. Blocking AI reading crawlers (bots that access your content in real time to answer a live query) removes you from AI search results entirely.
The setup we now recommend to every client:
- Allow AI reading bots (GPTBot, ClaudeBot, PerplexityBot, Googlebot-AI) in robots.txt
- Optionally restrict AI training bots separately if protecting content matters to you
- Add an llms.txt file to your domain root so AI tools understand your site structure, your key pages, and what you do
How long does it take to appear in AI search?
The honest answer is three to six months, often longer. AI knowledge bases are not updated in real time. After you publish content, it needs to be crawled and indexed by Google first, then picked up in the next AI knowledge base refresh. Google AI Overviews have a tighter relationship with Google's own index, so well-structured content that Google indexes quickly can appear in AI Overviews within a few weeks. For ChatGPT and Perplexity, plan for a longer runway. Regulated industries like legal services and financial services have additional E-E-A-T weight to navigate.
How do you maintain AI visibility over time?
Getting into AI search results is not a one-time task. Content that was accurate eighteen months ago may be outdated at the next refresh cycle. Content freshness matters as a signal in its own right. In practical terms:
- Reviewing and updating your most important pages at least twice a year
- Publishing new content at a regular cadence, even if volume is modest
- Monitoring brand search volume in Google Search Console as a proxy for AI recommendation activity
- Tracking direct traffic from AI tools that carry referral headers as a visibility signal
- Building external mentions through outreach, press, and community presence
The metrics we now lead every client report with are: tracked conversions via Google Tag Manager (phone clicks, form fills, email clicks), brand search trend, AI referral signals, and visibility in the Google results page.
Where to start
Start with the technical check. Look at your robots.txt and confirm AI reading bots are not blocked. Check your Cloudflare settings if you use it. Then audit your most important pages for E-E-A-T signals: named author, specific data points, external citations. Rephrase generic section headings as questions. If you want a second pair of eyes on this, book a discovery call.
By Chris McDowell, founder of BrisTechTonic. We help UK businesses get found by real customers, including the growing number who now start their search in AI tools.