You make your website easy for AI to quote by answering the questions your buyers actually ask, in short, self-contained answers, on the pages where those questions come up, and then labelling the facts so a machine knows exactly what they are. None of that is new. It is good SEO practice with the dial turned up, because the reader deciding whether to repeat your words is now a machine that writes its own answer.

I closed the Digital Knowledge Exchange conference at Elland Road in Leeds on this topic on 14 October, and the questions from the room were sharper than the slides. So this guide follows the order the room asked in.

Why does AI quote some websites and not others?

AI quotes the pages that give it a clean, specific answer it can trust, and it names the source only when naming the source adds something. The way I explained it on stage is the radio. You hear song after song, and only some of the time does the DJ tell you who it was. AI is the same. It plays your content all day long, and only now and then does it say your name.

That matters commercially because fewer people now click through to check.

In the period Semrush measured, AI Mode was only around 1% of Google sessions, so read that as a direction of travel rather than today’s reality for every search. The bigger slice is AI Overviews, the summary that appears above the normal results: Conductor’s 2026 benchmarks found one on around 25% of the searches it tracked across more than 13,000 enterprise websites. And the pages those summaries cite are mostly ones that already rank well: Ahrefs found 76% of AI Overview citations come from pages in Google’s top 10.

That is not the death of search. Google still matters, and the basics still need doing. What changes is where the decision gets made. The person who does arrive has usually read the AI’s summary first, and has arrived because you were named in it.

What does an easy-to-quote page look like?

An easy-to-quote page answers one real question per block, in two or three sentences, with the answer in the first sentence. That is the shape an AI tool can lift word for word. Long, clever paragraphs that build to a point get skipped for a competitor who just said it.

In practice, that means a bank of three to five questions at the bottom of your homepage, every service page and every product page. Not the same five questions repeated everywhere. “How much is delivery?” on every page is wallpaper. The questions on a service page should be about that service, the ones that come up in the first phone call or the first email about it.

The research backs the shape. In a 2024 study by researchers at Princeton and IIT Delhi, presented at KDD, adding quotations, statistics and cited sources to a page raised its visibility in AI-generated answers by up to around 40%, while keyword stuffing made it worse. Specific beats clever.

How do you write FAQs for a machine without sounding like one?

You write them for the customer and let the machine overhear. The first question I got at Elland Road, before I had reached slide two, was how obvious the FAQs need to be and how much you should write for the machine. The answer is that the machine is copying your customer. The people typing into ChatGPT are typing the same questions they would ask you on the phone, only longer and more specific.

So start from their language, not yours. The example I always use is a school. If a school makes an enquiry, it expects answers that talk school: DBS checks, governors, community engagement. If your page talks about “bespoke solutions for the education sector”, you have answered a question nobody asked. The same goes for size. A two-person business does not talk about its senior leadership team. If your buyers are small, write small.

One attendee asked whether it was fine to tuck an FAQ section at the bottom of a blog post. It is, as long as the questions belong to that post and the answers stand on their own if lifted out.

Two things to avoid. Do not stuff keywords into the answers; that is the old SEO habit and it reads badly to both audiences. And do not invent questions to fill the bank. Three real ones beat five padded ones.

What does AI not know about your business?

AI does not know the things you take for granted, because you have never written them down. I explain this with the Great British Bake Off. Imagine the programme was only the final scene, the finished cake and the tasting. You would miss everything that makes it worth watching: how they got there, what went wrong, what they learnt, the ingredients that made the difference.

Most websites are the final scene. The experience, the failures, the things you learnt, the awards, the reason a customer trusts you: none of it is on the page, so none of it is in the answer.

If it is only in your head, AI cannot quote it. If it is only in a logo, AI cannot read it.

That second half came from a good question in the room. Someone asked whether a Chamber of Commerce logo on every page would help. For the human, yes: it is a quick trust signal. For the machine, a sentence on your About page saying you are a member, since when, and why, does far more than the badge. Write the narrative; keep the badge for the people.

So go through your About page, your homepage and your main service pages and pepper in the specifics: years in the trade, how you learnt it, things you tried that failed, the kinds of customers you have done it for. Be explicit. It feels like stating the obvious because, to you, it is.

What is schema, in one sentence?

Which labels matter depends on what you sell. This is where I would start.

Schema type names as published at schema.org, checked October 2026. On a phone, scroll the table sideways.

A related one came up in the room too. An attendee who sells one product in several colours asked whether the near-identical pages for each colour would count against them. The fix is a canonical tag, a line of code on each colour page that says “the main version of this is over there”. Same principle as schema: label it, so the machine does not have to guess.

Here is what it looks like when it is built in from the start. When we rebuilt the website for a Yorkshire B2B technology company early in 2026, every template carried the labelling as standard: summary blocks at the top of pages, expanded organisation schema covering the area served and what the business is known for, author bios linked from the service pages, and copy written to a reading age a 14-year-old could follow. In the first quarter after launch, clicks on its main branded local search rose 115%. That figure includes a rebrand settling in, so I would not credit all of it to the labelling. What I would say is that a site built this way gives the machine nothing to guess about.

How do you know if it is working?

You ask the machine. Pick the question a buyer would type, something like “who is the best [your trade] in [your town]”, run it in ChatGPT, Google’s AI Mode and Perplexity, and then do not stop at the answer. Ask it why it named who it named, who else it considered, and why you were not in the list. At Elland Road I asked the room to try it on their phones. Within a minute, one attendee found the AI had confused their business with another on a near-identical web address. That is a labelling problem, and it is fixable.

We wrote up the full routine, with the exact prompts and what to look for, in how to check what AI says about your business. If the check turns up technical gaps, our 20 urgent SEO actions for the AI search era covers the fixes. Run it before you change anything, then again six to eight weeks after, and keep the answers. AI results move around, so one run tells you very little; the pattern across several runs is what counts.

FAQ

How many FAQs should each page have? Three to five, unique to that page. More than that and the questions start to get generic; fewer and you have not covered the real conversations. Put shared questions like delivery and returns on one FAQ page.

Does FAQ schema still get me a special listing in Google? No, Google no longer shows FAQ rich results for most websites. Mark up your FAQs anyway as housekeeping, but the visible, well-written answers are what AI tools lift and what people read.

Do I need to rewrite my whole website? Usually not. It is a refinement job, not a rebuild. Start with your homepage and your three highest-value service or product pages, add the questions and the specifics, and work outwards.

Is this different from SEO? It is SEO with a new reader in mind. Answering real questions clearly, showing your experience and labelling your data have always helped in Google. AI tools reward the same habits, and punish their absence more quickly.

Jonny Ross founded Fleek Marketing and has spent 27 years in search and visibility work, starting with his own online sunglasses business in 1999. He has hosted and spoken at the Digital Knowledge Exchange conference for UMi three times.