The AEO Checklist: 12 On-Page Moves to Get Cited by AI in 2026
Sajad Saleem
Co-founder of Ampliflow. Builds AI automation, websites, SEO/AEO, and growth systems for UK SMEs.
A plumber in Coventry types his own trade into ChatGPT — "best emergency plumber near me" — and watches it recommend three rivals, not him. His website is fine. It ranks on Google. The model simply could not find a sentence on it worth repeating. An AEO checklist is the fix: a short, ordered set of on-page moves that make your content easy for an AI engine to find, trust, lift and attribute. This is that checklist — twelve moves, in priority order, for UK businesses in 2026.
Last updated: June 2026 · Built for UK SMEs · Tactics verified against each AI provider's published crawler documentation and the GEO research cited below
TL;DR:
- The checklist is the unlock. AI engines quote structured, liftable formats — so an ordered checklist is exactly the kind of page they cite. Practise what you list.
- Being cited is not a bigger version of ranking. It is a different unit of work: you are no longer winning a position, you are winning a sentence.
- Most of the leverage sits in four free moves you can finish this week — let the crawlers in, lead with the answer, one idea per paragraph, ship the schema. Budget is not the constraint. Discipline is.
- The single biggest differentiator is a number nobody else has. Publish your own data and you become the source the model has no alternative but to quote.
- Domain authority counts for less here than in classic search. The clearest, best-sourced page wins — which is why a focused UK firm can out-cite a national one.
A checklist is the format AI lifts — so we made the definitive one
There is a small irony worth naming. The content AI engines quote most readily is the content easiest to lift in one clean block: ordered steps, direct answers, tables, definitions. A checklist is that format. Write a good one and you are not only describing how to get cited — you are demonstrating it.
The twelve moves fall into three tiers. Tier one is the floor — skip it and nothing else matters, because the engine cannot read your page or cannot extract a clean answer. Tier two is where most businesses pull ahead, since almost no UK competitor does these well. Tier three is the quiet infrastructure underneath.
An AI engine does not rank your page. It reads it and decides whether a single sentence is worth repeating with your name on it. Everything below is the work of making that sentence easy to find.
The 12 moves at a glance
| # | Move | Why it gets you cited | Priority |
|---|---|---|---|
| 1 | Let the AI crawlers in | If the bot is blocked, you cannot be quoted — full stop | Tier 1 · Floor |
| 2 | Lead with the answer | The model lifts a self-contained 40–60 word reply; buried answers get skipped | Tier 1 · Floor |
| 3 | One idea per paragraph | The paragraph is the unit an LLM extracts; it must stand alone | Tier 1 · Floor |
| 4 | Ship the schema stack + a real FAQ | Structured data tells the model what the page is and how to trust it | Tier 1 · Floor |
| 5 | Publish original data | Your own numbers are the one thing rivals cannot copy or out-source | Tier 2 · Differentiator |
| 6 | Build a comparison table | Tables are liftable wholesale and answer "X vs Y" prompts directly | Tier 2 · Differentiator |
| 7 | Define every term plainly | Models quote clean "X is …" sentences and make you the source of record | Tier 2 · Differentiator |
| 8 | Keep pages genuinely fresh | Recently maintained pages are favoured, and the date is a trust cue | Tier 2 · Differentiator |
| 9 | Make your entity unmistakable | Consistent name, address and schema let the model resolve who you are | Tier 2 · Differentiator |
| 10 | Publish an llms.txt | Low-cost hygiene; may help as adoption grows (consumption unconfirmed) | Tier 3 · Infrastructure |
| 11 | Cite your own sources | Pages that attribute credible sources get cited more themselves | Tier 3 · Infrastructure |
| 12 | Keep the page fast | Slow pages are read less and cited less | Tier 3 · Infrastructure |
Tier 1 — The floor: do these, or nothing else gets read
1. Let the AI crawlers in
The dullest reason a capable UK site is never cited: its robots.txt blocks the bots that would quote it. Open the file and confirm it allows GPTBot, OAI-SearchBot and ChatGPT-User (ChatGPT); ClaudeBot, Claude-SearchBot and Claude-User (Claude); PerplexityBot (Perplexity); and Google-Extended, which governs whether Google may use your content in Gemini and AI Overviews. It is a one-line change. Access is binary — a blocked page has a citation probability of zero, however good it is.
2. Lead with the answer
Open every page, and every major section, with a self-contained reply to its exact question — roughly 40 to 60 words, before any wind-up. A model lifting an answer wants a passage it can quote whole, with no preceding context. Bury it three paragraphs down and the engine moves to a competitor who put theirs first. This is the instinct behind LLM SEO: win a sentence, not a position. Answer, then explain. Never the reverse.
3. Give each paragraph one liftable idea
A model does not read your page top to bottom; it extracts paragraphs, and the paragraph is the unit it lifts. So each must make sense pulled out of context. Keep them short, one idea each, with no connective tissue that only works in sequence — cut "as we saw above"; the model sees this block, alone. One clean idea is quotable without risk. A four-idea wall gets skipped.
4. Ship the schema stack and a real FAQ
Structured data tells a machine what your page is. Mark up content with Article, FAQPage and BreadcrumbList schema; pages carrying multiple well-formed types are parsed and trusted more readily than bare HTML, and Gemini — grounded in Google's index — benefits most. Pair it with a genuine FAQ, because question-and-answer formatting maps onto how people prompt an AI and is among the most readily quoted structures. Use the questions customers actually ask.
Tier 2 — The differentiators: what separates a cited page from an ignored one
5. Publish a number no one else has
This is the move that pulls you ahead of every rival hiding behind vague copy. The team led by Aggarwal at Princeton and IIT Delhi (arXiv:2311.09735, presented at ACM SIGKDD 2024) tested nine ways to optimise for generative engines; the methods that lifted visibility most were adding statistics, citing sources and quoting experts. Your own data is the strongest form of all three — a small benchmark, a client result with a real figure, a sector survey. We did it with our 45-query citation test. Almost no UK competitor publishes original data. That is the gap.
6. Build the comparison table the model wants to lift
When someone asks an AI "AEO vs SEO" or "ChatGPT or Perplexity for my business", the engine reaches for a structured comparison — and if your page holds a clean one, you are the obvious thing to quote. Tables lift wholesale, rows and all. Build the comparison your buyer is weighing: traditional SEO against AEO, one platform against another, your service tiers side by side. Keep each cell short and self-contained, because the model may quote a single row.
7. Define every term in one clean sentence
Give every key term a crisp "X is …" sentence. Models lift definitions wholesale, and a clean one makes you the source of record — the page the engine returns to whenever the term arises. Define plainly and early, in the customer's words: "Answer engine optimisation is structuring content so AI engines quote it directly" beats two hedged paragraphs that circle the point. A cross-linked glossary compounds the signal.
8. Keep the page genuinely fresh
Recently maintained pages earn more citations, and the "last updated" date is itself a trust cue the model reads. Stale content is quietly discounted. The honest version matters: show a real date and make real edits behind it — refresh a statistic, add a question, sharpen a definition. A date stamp over untouched content is a tell, not a tactic. Put your most important pages on a genuine review cycle.
9. Make your business impossible to confuse with another
An engine can only cite you for "AI consultancy near Birmingham" if it is sure who you are. That confidence comes from entity consistency: the same name, address and founders, stated identically across your site, directories and social profiles, reinforced with Organization and LocalBusiness schema and sameAs links. For UK firms this is a quiet advantage, because so few competitors make their locality explicit and machine-readable. A clear entity is easier to recommend than a fuzzy one.
Tier 3 — The infrastructure: the quiet foundations under everything
10. Publish an llms.txt
An llms.txt file is a plain-text map of your most important pages — each with a title and short excerpt — placed at your root for AI systems to read. It hands a model a clean index of what you offer, instead of leaving it to infer your structure from the navigation. Treat it as cheap, no-downside hygiene, not a citation lever: no major AI provider has yet confirmed it reads the file, so the AEO payoff is unproven. Publish one anyway — it costs an afternoon, leaves you ready if adoption grows, and gives you a tidy map of your own business either way. Keep it current as you add key pages.
11. Cite your own sources
Counter-intuitively, pages that link out to credible sources get cited more themselves. Attribution places you inside a trustworthy chain of information, and a model registers that you do not assert numbers from nowhere. So name your sources — link the study, credit the data, point to the original. It buys a trust signal unsourced rivals cannot fake. The engines reward the behaviour they themselves depend on.
12. Keep the page fast
Slow pages are read less and cited less. An engine fetching your content under time pressure favours a page that loads cleanly, and a poor mobile experience compounds the loss. This is where AEO and ordinary web quality become one conversation: a fast, well-built website serves human buyers and AI crawlers with identical work. The best foundation for getting cited is simply a good site, kept quick.
Where to start if you only have an afternoon
Do not attempt twelve at once. The order is the point. Spend the first hour on tier one, because it is free and gates everything else: open your robots.txt and let the crawlers in, then take your three most important pages and rewrite their openings as direct, self-contained answers. That alone moves more pages from invisible to quotable than anything further down the list.
Then choose one differentiator and commit to it — almost always number five. One real figure of your own, published with its method and source, does more for your citations than a month of tidying. The businesses that win AI search are not the ones doing all twelve flawlessly. They are the ones who did the floor, then published something true that no one else could.
Frequently asked questions
What is an AEO checklist?
An AEO checklist is an ordered set of on-page actions that make your content easy for AI engines — ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews — to find, extract, trust and attribute. It runs from access through extractability to differentiation and the infrastructure beneath. The goal is not a higher ranking; it is to be the source the model quotes by name.
How long does AEO take to show results?
Weeks to a few months, not days. Models must re-crawl your updated pages, and live-search behaviour shifts gradually as your content is re-evaluated. Freshly published original data tends to be picked up fastest. Measure by asking the engines your customers' questions once a month and reading the trend, rather than over-reading any single answer.
Do I need an agency to do this?
No. The whole of tier one is free and within reach of any owner willing to edit a robots.txt, rewrite a few opening paragraphs and add structured data. The items that genuinely benefit from help are the differentiators — designing an original-data study, or building schema cleanly at scale. Start with what you can do yourself this week.
Which item matters most if I can only do one thing?
Two, and they are inseparable: let the crawlers in, and lead every page with a self-contained answer. Access plus extractability is the floor — without both, nothing else can surface. Once that is in place, publishing original data is the strongest single differentiator, because almost no UK competitor offers numbers of their own.
Is AEO a replacement for SEO?
No. AEO sits on top of solid SEO; it does not replace it. A crawlable, fast, well-structured site is the precondition for both, and Gemini grounds its answers in Google's index, so strong classic SEO directly improves your odds of being cited there. Treat them as complementary layers of one programme — the deeper split is covered in AEO vs SEO.
Should I really let AI crawlers read my content?
For almost every small business chasing visibility, yes. Blocking the AI crawlers removes any chance of being cited in the channel where your buyers increasingly start. The narrow exception is genuinely proprietary material you never want surfaced — keep that behind a login or expressly disallowed, and let the pages you want quoted stay open.
Does a UK location help me get cited?
It helps, when you state it clearly. Models resolve businesses by entity signals, so a consistent name, address and founder set — reinforced with LocalBusiness schema — makes you easier to cite for region-specific questions. "Accountant near Solihull" or "AI consultancy in the Midlands" is winnable precisely because so few firms make their locality explicit and machine-readable.
Related reading
- ↑ From SEO to AEO: How UK Businesses Can Dominate AI-Powered Search — the pillar guide and the full framework
- ↔ What Is Answer Engine Optimisation (AEO)? — the foundational definition
- ↔ LLM SEO: Optimise for ChatGPT, Claude & Gemini — the model-by-model cut of the same work
- ↔ Generative Engine Optimisation (GEO): The Complete UK Guide — the research and the deep tactics
- ↔ How to Get Cited by ChatGPT, Gemini and Perplexity — platform-by-platform tactics
- ↔ What AI Search Actually Cites for UK Business Queries (2026 Data) — our original 45-query benchmark
- ↔ How to Show Up in Google AI Overviews — the checklist applied to Google's answer box
- ↔ How to Get Cited by Perplexity AI — the checklist applied to Perplexity
- ↔ Does Your Business Need an llms.txt File? — checklist item #10, in depth
Run the checklist against your own site
Open ChatGPT tonight and ask it the question your best customer would ask. If it names a competitor and not you, you have just found this quarter's most important growth project — and the list above is the order in which to fix it.
Get a free AI-visibility audit → — we run your key buyer questions through ChatGPT, Claude, Gemini and Perplexity, score your pages against all twelve moves, and hand you the prioritised fixes that move you into the answer. Forty-five minutes, no commitment.
A checklist earns nothing by being read — only by being worked. Twelve moves, three tiers, one order: start at item one this afternoon, and let each fix compound into the next.