Someone asked ChatGPT last week: "who are the best AI automation agencies for UK dental practices." A name came back. A real company, with a one-line summary lifted — almost verbatim — from their homepage.

That company didn't buy a SERP slot. They didn't rank #1 on Google for that query. They ranked inside the answer.

That's the shift. And most SME websites are still optimising for the 2018 version of search.

What AEO and GEO actually mean

AEO — Answer Engine Optimization — is making your content the source an LLM cites when someone asks a question adjacent to what you sell. ChatGPT, Claude, Perplexity, Gemini, Google's AI Overviews.

GEO — Generative Engine Optimization — is the more general term for the same idea: optimising for generative search interfaces, not ranked blue-link lists.

They're not replacements for SEO. They're an extension of it. The same site that ranks in Google's classic SERP often shows up in AI answers too — but only if you've done two specific things most SME sites still haven't.

Shift 1: stop writing for keywords, start writing for questions

The old SEO model: pick a keyword ("dental practice software"), stuff it into a page titled around it, build backlinks. The LLM model: someone asks a question, and the model synthesises an answer from the cleanest, most quotable snippet it can find.

What LLMs reward:

  • Short, declarative answers placed near the top of a page, with a clear H2 that is literally the question.
  • Structured lists (3–7 items) the model can lift wholesale.
  • Named experts or case studies with concrete numbers — models love specific numbers because they reduce hallucination risk.

The classic 2,500-word "ultimate guide" that wanders through 16 H2s before answering the question? Terrible for LLM citations. The model gives up halfway through and picks a competitor.

Rewrite your top-10 conversion pages around a single question each. Answer it in the first 80 words. Back it up with evidence underneath. You'll lose nothing in Google ranking (often gain it — Google also rewards concise, question-aligned content now) and gain enormous lift in AI citation rate.

Shift 2: structured data, for real this time

Schema.org JSON-LD has been table stakes for SEO since 2014. The difference now: LLMs read it during their crawl, and use it as a reliability signal.

The schemas that matter most for SME B2B sites:

  • Organization with full contactPoint and sameAs — lets the model answer "who are they?"
  • Service with a proper offers array containing Offer with price and priceCurrency — lets the model answer "how much does it cost?" accurately.
  • FAQPage wrapping your FAQ section — a favourite for direct LLM citation.
  • Article on every blog post with author, datePublished, and mainEntityOfPage.

We add all four to every client site now as part of standard SEO work. It's an hour of implementation for a site that's otherwise built correctly. We see it measurably in the referral logs from Perplexity and Claude inside 4–6 weeks.

What to actually do before Q3

A pragmatic checklist:

  1. Audit your three highest-value pages — the ones that, if someone landed on them from an LLM answer, would convert. For most SMEs: service page, pricing page, contact page.
  2. Restructure each around one question. H1 is the question. First paragraph (under 80 words) is the answer. Supporting evidence follows.
  3. Add Organization, Service, and FAQPage JSON-LD to the site. Validate with Google's Rich Results Test — if it passes there, LLMs will parse it fine.
  4. Put your prices in machine-readable form. A table with real numbers beats a paragraph that says "prices start from around £X." Models quote what they can parse.
  5. Name your clients (or their sectors) in case studies with concrete metrics. +312% enquiries over 11 weeks is quotable. We helped them grow significantly is not.
  6. Track AI-referral traffic separately. In Plausible or GA4, segment by referrer. The rising ones — perplexity.ai, chat.openai.com, claude.ai — are the new source-of-truth channel.

The one that will annoy your SEO agency

The old SEO playbook valued length and keyword density. The new playbook values citability. Those are different, and sometimes in conflict.

A 500-word page that's structured as a question and a cited answer will almost always outperform a 3,000-word "pillar page" in AI citations. Your SEO agency may still measure success by word count and internal links. That metric is increasingly disconnected from revenue.

If you're spending £1,500+/mo on SEO and your agency hasn't mentioned AEO, GEO, or schema validation to you in the last three months, you're paying for 2022's playbook.

Where this goes

Google's AI Overviews now show on roughly 30% of UK queries and rising. Perplexity's user base is growing 30% month-over-month. Claude's claude.ai web interface quietly added web browsing in Q1.

Search hasn't died. It's just stopped being a list of links and started being an answer — one that cites two or three sources. Being one of those sources is the SEO job for 2026. Everything else is commentary.

If your site still ranks well on Google but you're not seeing referral traffic from Perplexity, ChatGPT, or Claude, that's a fixable gap. Usually a 30-day SEO sprint.