A growing share of your customers no longer type keywords into a search box and scan ten blue links. They ask ChatGPT, Copilot, Claude or Perplexity a full question, get a paragraph back, and act on it. Even Google now opens many results pages with an AI Overview. When the answer is a summary rather than a list, the question stops being only “do we rank?” and becomes “does the machine writing the summary describe us accurately, or at all?”
Answer engine optimisation (AEO) is the practice of improving that outcome. It has already attracted its share of snake oil, so here is the honest version: most of AEO is traditional technical SEO done properly, plus a layer of plain-language clarity and machine-readable facts. This is part of our web development and SEO work, and everything below is practised on this website.
How assistants learn about your business
- Training data. Models carry a snapshot of the public web that is months or years old. You influence it slowly, by having accurate content that others reference over time.
- Live retrieval. The one that matters most. When someone asks about your industry or your business, assistants increasingly run a web search, fetch a handful of pages, and quote what they read. Improvements here pay off in days, not years.
- Machine-readable declarations. Sitemaps, structured data, and emerging files such as llms.txt that summarise your site for machines.
You cannot buy a ranking inside a model. What you can control is what an assistant finds at the moment it looks you up.
Traditional SEO still carries most of the weight
AI search tools lean heavily on conventional web indexes to decide which pages to fetch. If your site is slow, broken on mobile, thin on content or invisible to crawlers, you are invisible to assistants too. The foundations have not changed: fast pages served over HTTPS, clean HTML with real headings, one clear topic per page, and working internal links. Performance and security remain the cheapest wins.
Write pages a machine can quote
Assistants quote text, so give them text worth quoting.
- Answer real questions in plain sentences. State who you are, where you operate, what you do and who it is for, near the top of the page, not buried under slogans. A sentence like “we provide X for Y in Z” is quotable; “empowering synergy” is not.
- Keep facts consistent everywhere. Business name, location, phone and services should match across every page and profile. Machines cross-check, and inconsistency reads as unreliability.
- Use FAQ sections. They mirror the way people phrase questions to assistants, and they concentrate answers in a liftable form.
- Add structured data. JSON-LD types such as Organization, Service and FAQPage turn your key facts into fields a machine can read directly instead of inferring from prose.
llms.txt and an explicit crawler policy
llms.txt is a proposed convention: a markdown summary of your site, at a known URL, written for language models. Adoption across the industry is still uneven, but publishing one costs almost nothing and states your facts in exactly the form assistants consume. We publish our own, alongside an ai.txt declaring how AI systems may use the content.
The other half is robots.txt. Each major AI provider publishes the names of its crawlers, and most separate training crawlers from search crawlers, so you can allow one and not the other. Blocking everything protects content you never want summarised, at the price of being absent from answers your competitors will appear in. For a business that wants to be found, allowing retrieval is usually the right call. Either way, make it a decision rather than a default, and keep genuinely private material off the public web entirely instead of trusting a crawler directive to hide it.
What to do this quarter
- Ask two or three AI assistants what they know about your business and your main service. Note what is wrong, missing or stale; that is your work list.
- Fix the foundations: page speed, mobile rendering, crawlability, and structured data on your key pages.
- Publish the facts you want quoted, as plain-language service pages and FAQs, then add llms.txt and set an explicit AI crawler policy in robots.txt.
Search is becoming a conversation your customers have with a machine, and the businesses described accurately in that conversation will win the enquiry before a website is ever opened. If you would like yours to be one of them, book a consultation.