The Content That AI Actually Uses
Ask ChatGPT which accountant to use in Leeds, or how long a house sale usually takes, and you’ll get a neat, confident answer in seconds. It feels like the AI has read the whole internet and made up its mind.
It hasn’t. What it’s actually done is pull a handful of short passages from a handful of sources and stitch them together. A sentence from one site. A price range from another. A review from somewhere else.
That changes what “good content” means. Most business websites are written to rank on Google, or to impress someone who’s browsing. Very little of it is written to be quoted. And if your content can’t be quoted, it can’t be used.
So here’s what AI tools actually pull from, what they tend to ignore, and what you can do about it without rebuilding your whole site.
The short version: AI search tools use content that answers a specific question clearly, in a passage that makes sense on its own. They favour concrete facts like prices, timelines and processes, honest comparisons, consistent information about who you are, and what other websites say about you. Vague brand copy and long guides with the answer buried deep tend to get skipped.
How AI tools decide what to use
You don’t need the technical detail, but the basic process helps. When someone asks a question, tools like ChatGPT, Perplexity and Google’s AI Overviews search for relevant content, break pages into smaller chunks, and pick the chunks that answer the question best. Then they write a response based on those.
The important word there is chunks. The AI isn’t judging your page as a whole. It’s judging individual paragraphs. A brilliant 3,000-word guide can lose out to a 60-word answer on a much smaller site, simply because the smaller site got to the point.
AI tools also look for confirmation. If your website says something and three other sites say the same, that’s a fact the AI can use with confidence. If you’re the only one saying it, or other sources contradict you, it’s a riskier bet, and the AI is more likely to look elsewhere.
We’ve covered more on how AI decides which businesses to recommend in a separate post.
The five types of content AI actually uses
1. Direct answers to specific questions
This is the big one. AI tools are answering questions, so they look for content that answers questions. Not content that’s loosely about the topic. Content that actually answers it.
Picture two conveyancing pages. One opens with three paragraphs about the firm’s history and its commitment to client care. The other opens with: “Most of the freehold purchases we handle complete within three to four months of an offer being accepted. Leasehold and new-build purchases tend to take longer.”
The second one gets quoted. The first gets scrolled past, by people and machines alike.
Quick fix: List the ten questions your customers ask most often. Then check whether your site answers each one in the first sentence or two of a relevant page.
2. Concrete facts
AI tools love specifics because specifics are useful. Prices, timelines, locations, process steps, who you work with, what’s included.
Plenty of businesses hold this back. A dental practice might not list Invisalign prices because every case is different. That’s fair enough, but it means when someone asks an AI how much Invisalign costs in their area, the answer gets built from a competitor who was willing to give a range.
You don’t have to publish a full price list. “Treatment usually costs between £X and £Y, depending on how complex your case is” is enough to be useful.
Quick fix: Find the facts you currently hide behind “get in touch” and see which ones you could give a range or a typical example for.
3. Honest comparisons
A huge number of AI searches are comparisons. “Xero or QuickBooks for a small business?” “Fixed fee or hourly rate for a divorce solicitor?” “Which CRM suits a ten-person sales team?”
The content that gets used here is content that genuinely helps someone choose. That includes being upfront about who you’re not right for. A SaaS company that says “we’re built for teams of 20 to 200, and if you’re a sole trader there are simpler, cheaper options” is giving the AI something it can use when matching products to people.
It feels counterintuitive. But being clear about fit makes you easier to recommend to the right people.
Quick fix: Write one page comparing the main options your customers weigh up, including the ones that aren’t you.
4. What other websites say about you
This is the part most businesses miss, because it doesn’t live on their website at all.
When someone asks an AI for a recommendation, like “best family law firms in Leeds” or “good accountants for contractors”, the answer leans heavily on third-party sources. Review sites, directories, industry roundups, “Top 10” lists, Reddit threads, local press. Your own website saying you’re great doesn’t count for much. Other people saying it counts for a lot.
Quick fix: Search for your business name alongside your main service and see which third-party sites come up. Check you’re listed on the directories that matter in your sector, and that your reviews are reasonably recent.
5. Clear, consistent information about who you are
AI tools need to understand what your business is before they can recommend it. Your name, what you do, where you do it and who you do it for all need to line up, everywhere.
It sounds obvious. But it’s really common for a business to be described one way on its website, another way on LinkedIn, and a third way on an old directory listing it forgot existed. An accountancy firm might be “chartered accountants” on its own site, “financial services” on one directory and “bookkeeping” on another. When the descriptions don’t match, the AI can’t be sure what you are, so it’s less likely to put you forward.
Quick fix: Pick one clear description of your business and update every profile you control to match it.
Content AI tends to skip
- Vague brand copy. “We deliver tailored solutions that drive results” doesn’t answer anything. There’s nothing in it to quote.
- “We do everything” positioning. If you call yourself a full-service, one-stop shop, the AI has no clear idea when to recommend you. Specific beats broad.
- Answers buried in long guides. If the useful bit is in paragraph nine, it’s up against sites that put it in paragraph one.
- Information that’s locked away. PDFs, text inside images, gated downloads and content that only appears after a click. If it’s awkward for a machine to read, it often doesn’t get read.
Before and after: one paragraph rewritten
Here’s a fairly typical service paragraph from a financial planning firm:
At [Firm], we believe every client deserves a personalised approach. Our experienced team takes the time to understand your goals and provides holistic advice to help you achieve financial peace of mind.
There’s nothing wrong with it, exactly. But if someone asks an AI “do I need a financial adviser for retirement planning?”, there’s nothing here it can use.
Here’s the same paragraph rewritten:
[Firm] is an independent financial planning firm in Chester that specialises in retirement planning for people aged 50 and over. We help clients work out when they can afford to retire, how to draw their pensions tax-efficiently, and how much they can safely spend each year. The first consultation is free, and ongoing advice is a fixed annual fee rather than a percentage of your investments.
It’s still friendly and easy to read. But now it says what the firm is, where it’s based, who it helps, what it actually does and how it charges. Any one of those sentences could be lifted into an AI answer and still make sense.
You don’t need separate “AI content”
When people first hear about this, there’s a temptation to create a separate set of pages written for AI. Don’t. You’ll split your effort, risk duplicate content, and end up with two weaker pages instead of one strong one.
The better move is to improve the pages you already have, so they work for people, Google and AI tools at the same time. In practice that usually means:
- Opening key pages with a clear, direct answer
- Adding an FAQ section that answers real customer questions in a sentence or two each
- Swapping vague claims for specific facts
- Making sure your business details match everywhere they appear
Most of this is just good SEO done properly. The bar for clarity has gone up, that’s all.
A five-minute check for your own site
Open your most important service page and ask yourself:
- Could someone copy any single paragraph and have it make sense on its own?
- Does the page give at least one concrete fact, like a price range, a timeline or a process step?
- Does it answer the question a customer would actually type, within the first few lines?
- Does your business description match across your website, Google Business Profile and LinkedIn?
- Are you mentioned on any third-party sites that someone in your sector would trust?
If you answered no to two or more, there’s probably a quick win waiting.
FAQs
Does AI search use the same content as Google rankings?
Partly. Google’s AI Overviews often draw on pages that already rank well, and ChatGPT and Perplexity use search results too. But ranking on its own isn’t enough. A page can rank well and still not get quoted if the answer is hard to pull out.
How long does it take for content changes to show up in AI answers?
It varies. Tools that search the web in real time, like Perplexity, can pick up changes fairly quickly once a page has been re-crawled. Answers that rely on what a model learned during training take much longer to shift. Third-party mentions build up over months rather than weeks.
Do I need to rewrite my whole website?
No. Start with your most important service pages and the questions customers ask most. A few focused changes to key pages usually make more difference than a full rewrite.
Can I see whether AI tools mention my business?
Yes. The simplest way is to ask them the questions your customers would ask and see who comes up. For ongoing tracking, AI visibility tools can run a set list of prompts regularly and report on how often you appear.
Curious what AI says about you?
We check how businesses show up in ChatGPT, Perplexity and Google’s AI answers as part of our SEO work. If you’d like to know where you stand, send us your company name and we’ll tell you what AI tools are currently saying about you, and who they’re recommending instead.
You can see how this worked for Ruler Analytics in our case study.