A growing share of your future customers will never see a results page. They'll ask ChatGPT which agency to shortlist, ask Gemini what a fair price looks like, ask Claude to compare three suppliers - and act on the answer. Google's own AI Overviews now answer many searches before a single blue link earns a click.
This isn't a reason to panic about SEO. It's a reason to understand a simple shift: search engines ranked pages; answer engines quote sources. Your job used to be earning a position. Now it's also being quotable.
We've spent this year applying that to our own site and our clients' - here's what actually moves the needle.
How AI assistants choose who to recommend
When an assistant answers a commercial question, it draws on its training data, live web retrieval, and structured signals - reviews, directories, schema. Across all three, the same properties win:
- Verifiable specifics. Concrete prices, timeframes, locations and numbers beat adjectives. "From £3,000, delivered in six weeks, based in Hampshire" is quotable; "bespoke solutions for ambitious brands" is noise an AI skips.
- Answer-shaped content. Pages that pose a real question and answer it directly in the first paragraph get lifted into responses. Pages that bury the answer under 900 words of throat-clearing don't.
- Third-party corroboration. Reviews, directory listings and mentions on sites that aren't yours. Assistants are trained to avoid recommending on self-praise alone.
- Clear provenance. A named author, a real organisation, consistent details everywhere. Ambiguity reads as risk, and assistants route around risk.
The seven-signal playbook
1. Publish definitive answer pages
Take the five questions your buyers actually ask - usually cost, timeframe, comparison, "is it worth it", "how to choose" - and give each a page whose H1 is the question, with the direct answer in the opening paragraph and real numbers throughout. This single format accounts for most of the AI citations we see. It also converts humans, because humans hate throat-clearing too. Our piece on what business automation actually costs is built to that shape, if you want the format rather than the theory.
2. Add FAQ and organisation schema
Structured data is how you hand your facts to machines on a plate: FAQPage markup on every Q&A block, Organisation schema with your real details, Service schema on service pages, Article schema with authors on posts. It's an afternoon of technical work that most UK SMEs simply haven't done.
3. Treat Google reviews as AI food
Review count and sentiment feed AI local recommendations directly - ask any assistant for "best [trade] near [town]" and watch review language echo in the answer. If your profile is thin, that's a this-quarter priority; we've written the full system in our reviews guide.
4. Make your details boringly consistent
Same business name, address, phone and description on your site, Google Business Profile, Bing Places, and the directories that matter in your sector. Contradictions make an assistant hedge - and a hedging assistant recommends someone else. The local-search groundwork is the same work: our Winchester local search blueprint walks through it town by town.
5. Put names on your work
Author bylines with a sentence of credentials, photos, a real about page, founders visible. Models weight content with accountable humans behind it - anonymous corporate content is exactly what they've been trained to discount.
6. Add an llms.txt file
An emerging convention: a plain-text file at yoursite.com/llms.txt describing who you are, what you do, who you serve and how to engage - written for machine readers. It takes an hour, almost nobody in the UK has one, and it's a free chance to author the summary an AI holds of you. (Yes, ours is live.)
7. Earn mentions you don't control
Trade directories, local press, supplier pages, industry round-ups. One genuine third-party paragraph about your business is worth more to an answer engine than ten pages of your own copy, because it's corroboration. Slow to build, impossible to fake, increasingly decisive.
What stays exactly the same
Site speed, crawlability, genuinely useful content, real expertise - all of it still underpins everything above, because assistants retrieve from the same web Google indexes. AEO isn't a replacement discipline; it's a sharper standard of clarity applied to the SEO you should already be doing, and our small business SEO checklist is still the right starting point. The honest summary: stop writing to impress and start writing to be quoted.
How to check where you stand
Ten-minute audit, today: ask ChatGPT, Gemini and Claude - "best [your service] in [your county]", "how much does [your service] cost in the UK", "[your business name] reviews". Note who gets named, what gets cited, and what the assistants believe about you. That gap between what they say and what you'd want them to say? That's the work.
FAQ
Is AEO different from SEO?
Same foundations, different finish line. SEO earns a ranking a human clicks; AEO earns a citation an AI repeats. Structured data, direct answers, provenance and third-party proof are the delta.
Can you pay to appear in AI answers?
Not in the organic answers of the major assistants today. The way in is being the most quotable, corroborated source - which is why early movers are winning cheaply.
How do we measure AI visibility?
Manually prompt the major assistants monthly and log mentions; watch referral traffic from AI domains in analytics; track branded search volume, which rises as assistants name you. Imperfect, but directionally clear within a quarter.
Does this matter for a small local business?
Arguably more. Local "best X near me" questions are among the most-asked commercial prompts, and in most UK towns nobody has done this work yet. The window where seven signals make you the default answer is open now - it won't stay uncrowded.
Want this applied to your website?
George or Nicholas will record a personal walkthrough of your site within 24 hours.