01
I start from evidence.
Every recommendation in my manuals has a source you can open. I label vendor research as vendor research, because most studies in this field come from companies selling tools. I treat Google’s guidance as Google’s stated position. It tells you what Google says, which isn’t always how its systems behave. When I share results from client work, I call them client work. Where the evidence is thin, I say so. I’d rather leave a gap than fill it with a confident guess.
02
AI search sits on top of search.
When AI assistants look something up, they mostly find pages through search indexes. Google says a page has to be indexed and eligible to show a snippet before it can appear in AI Overviews or AI Mode1. Practitioners I trust say the same2, 3. A page still doesn’t need the top spot. Pages lower on the first page get pulled into Google’s AI Mode too3. So I treat search and AI visibility as one job. The foundations are the same. Measurement is what differs.
03
Match the question.
Buyers ask about the same things in recognisable ways. Pages whose headings say plainly what each section answers, in the buyer’s words, are easier for AI to match and quote. In Cyrus Shepard’s survey of more than 130 search experts, a close match between the question and the answer was among the highest-rated factors for being cited in Google’s AI4. That’s expert opinion. It’s also the cheapest change you can make. Most marketers skip it because it feels too simple. It’s where I’d start.
04
Focused pages answer better.
AI quotes passages. Gemini, for example, passes only a few hundred words of a page into its answer5. A section that answers one question clearly is easy to lift. A long page that wanders buries the answer. Coverage still matters. It comes from a set of related pages that link to each other, each with a clear job. I don’t build a page for every variation of a question. Overlapping pages compete with each other, so I combine them.
05
AI can’t quote what it can’t read.
The pages AI uses are the ones it can read cleanly: the answer near the top, specific facts and plain formatting. The main content needs to be in the page itself, where a crawler can see it. Structured data helps search engines understand a page. Google says no special markup is needed for its AI features1. Most businesses bury their answers in marketing copy. Clear writing comes first.
06
Clarity makes real expertise visible.
Authority still helps. Bigger, better-known sites get found more often. Once a page is in the running, a clear answer can beat a better-known page that buries its answer. Florence, a pet cremation business I’m launching, had only a coming-soon page and a blog. Its blog posts earned their first ChatGPT citation within 14 days.
Clarity only works on substance. For brands with real expertise, clear presentation makes that expertise easier to find and quote. For brands without it, no formatting closes the gap. The fix is the underlying work.
07
Your site owns the facts. Others decide how you compare.
Which sources AI uses depends on the question. In a test of US prompts, ChatGPT leaned on official sources, including brands’ own sites, for factual questions like price, features and company details. For comparisons, it leans on other sites: reviews, publishers and communities6. So your own pages need complete, accurate facts. You also need other people talking about you, because that’s where comparisons get decided. The assistants differ from each other in which sources they favour, too7. The manuals go deeper on both.
08
Being named matters more than being linked.
Much of AI search’s value never shows up as a visit. A buyer reads the answer and decides without landing on your site. A citation link and a recommendation are also different things. Semrush found 62% of AI citations didn’t name the brand behind the cited page8. Brand mentions across the web track far more closely with being named in Google’s AI Overviews than backlinks do9. So I measure presence: whether AI names you, how it describes you, and how that changes over weeks. The AI traffic in your analytics is a floor10.
09
This builds over time.
There’s no shortcut that lasts. I don’t use mass-produced pages, link schemes or tricks aimed at AI answers. Google says pages made to manipulate AI answers break its spam policies1, and recovering from a penalty can take months or years. The work compounds: clear pages, accurate facts, and a reputation others write about. I wrote the manuals to be worked through steadily and thoughtfully, by people who know their business.
Sources (10)
Show
- Google Search Central, Optimizing your website for generative AI features on Google Search. Updated 10 July 2026. Pages must be indexed and snippet-eligible. No special markup needed. Pages made to manipulate AI answers break spam policy.
- Lily Ray, Your GEO Strategy Might Be Destroying Your SEO. 18 March 2026. Content has to be indexed and ranking to reach the model.
- Dan Petrovic, In AI SEO #10 is the new #1. 9 November 2025. Any first-page ranking can feed Google’s AI Mode.
- Cyrus Shepard, Google AI Ranking Factors. 16 September 2026. Survey of more than 130 search experts. Crawl access and question-answer match rated highest.
- Dan Petrovic, SRO & Grounding Snippets. 1 March 2026. Gemini passes only a few hundred words of each page into its answer.
- Ahrefs (Louise Linehan; data Xibeijia Guan), “How ChatGPT Does Query Fan-Out (15 Million Queries Studied)”. 12 October 2026. Factual questions cite official sources. Comparison questions lean on other sites. Vendor research.
- Foundation × AirOps, The State of B2B AI Discovery 2026. Data June to September 2026. Assistants lean on different sources. Sponsored vendor research.
- Semrush (Margarita Loktionova, with Kevin Indig), Why 62% of AI citations don’t lead to brand mentions. 9 June 2026. Across 115 prompts on four AI tools, most citations didn’t name the cited brand. Vendor research.
- Ahrefs (Louise Linehan; data Xibeijia Guan), An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied). 26 May 2025. Brand mentions correlated more strongly with AI Overview visibility than backlinks. Vendor research.
- Aleyda Solís, A 3 Layer Framework to Measure AI Search Presence, Readiness and Business Impact. 26 August 2026. Observed AI referral traffic is the floor.