AI SEO Cheatsheet

The acronyms multiply faster than the tactics. AEO, GEO, LLMSEO, AImention, AAO. They mostly describe the same work. This cheatsheet untangles the terms, then gives you the one checklist that covers all of them, with links to the deep-dive cheatsheet for each piece.

Eight of these topics are also drawn out as full how-to cards: SEOSHOTS: 8 SEO infographics.

The AI search acronym map

What each term means, and how it relates to the others. Spoiler: they overlap heavily.

The honest take: these are not five disciplines. They are facets of one job, be the source AI trusts and cites. Pick a term, stop arguing about it, and do the work below. See why AEO is not separate from SEO.

The AI SEO checklist (covers AEO, GEO and LLMSEO)

One checklist for the whole umbrella. Each row links to the deep-dive cheatsheet.

Which AI cheatsheet for which job

How AI SEO is measured

Stop reporting AI SEO as referral clicks (they are tiny and lagging). Measure grounding from your server logs: AI Crawler hits mean a page is being ingested, AI Assistant hits mean it is being cited live. The flip from crawler-dominant to assistant-dominant is the indexed-to-cited conversion. Track it per URL plus your Bing AI citation share.

Rewrite a paragraph into AI-quotable triples with the free Semantic Triple Studio tool.

New to SEO? Start with the fundamental SEO cheatsheet first.

Part of all the free SEO cheatsheets.

How do you run an AI SEO content workflow without hallucinations?

The workflow that holds up at scale is five steps with a human at the first and the last: prepare the data, brief the model, draft, edit, verify. Most teams skip the first and last, which is where the hallucinations and the thin pages come from. This is the template I run, and it works the same for an affiliate site publishing 40 posts a month and a B2B SaaS blog publishing four.

Step 1: prepare your data before it touches a prompt

• Export your Search Console queries with the page each one currently ranks for, so the model knows what already exists and does not write a cannibalising page.
• Strip branded queries and dedupe near-variants; a model given 900 raw rows will invent patterns that are not there.
• Tag each cluster with intent (informational, comparison, transactional) in a column. Feed the file as CSV or a table, never as a paragraph of pasted keywords.
• Add your source material: product docs, first-party data, customer calls. The model can only be accurate about what you give it.

Step 2: the brief prompts (copy and paste)

Content brief: "Using only the attached query cluster and source notes, write a content brief for [topic]: the primary question, the six sub-questions in order of search demand, the entities that must appear, the unique data we hold, and the three pages on our site to link. Do not add facts that are not in the sources. Mark anything you are unsure about with [VERIFY]."

On-page set for a product page: "For the product page in the attached HTML, write a title tag under 60 characters, a meta description under 155 characters, and a four-question FAQ block. Every claim must come from the page content. Output the FAQ as HTML with h3 questions and one-paragraph answers. Do not produce FAQ schema; the page template handles markup."

Outline for a B2B post: "Draft an outline for [topic] aimed at [role] evaluating [category]. Question-form H2s ordered by demand, one line under each stating the answer. Flag every place where we need a screenshot, a number, or a quote from a customer."

Steps 3 and 4: draft, then a human edits

The model drafts against the brief. A human editor then does three things no prompt replaces: cuts what the sources do not support, adds the first-hand detail (the screenshot, the number, the opinion), and rewrites the opening so it answers the question in the first sentence. Budget 30 to 40 minutes of editing per 1,000 words; if the editor spends longer, the brief was weak.

Step 5: why the prompts hallucinate, and the verification gate

Prompts hallucinate when the model has no source and is asked for specifics anyway. Every fix is a version of "give it the source or forbid the specific": attach the document instead of describing it, instruct the model to quote with a URL for every statistic, tell it to write [VERIFY] rather than guess, and lower the temperature for factual tasks. Then the gate: every number, name and date in the draft is traced to a source before publish, and anything untraceable is cut. Run the finished page through the low quality content checker before it ships.

A note on AI internal-linking suggestions

Tools that suggest internal links from embeddings, including my own Link Weaver, are good at finding which pages relate and bad at choosing anchor text. Take the target suggestion, then write the anchor yourself as the target page's title or a natural phrase, and keep to three to seven body links per page. Repeating the exact-match keyword as every anchor is the over-optimisation pattern that reads as manipulation on WordPress blogs.

Frequently asked questions

What is AI SEO?

AI SEO is optimising your brand and content to be found and cited across AI search surfaces (ChatGPT, Gemini, Google AI Overviews, Perplexity, Copilot). It is the umbrella that covers AEO, GEO and LLMSEO, which are facets of the same goal: being the source AI trusts and cites.

What is the difference between AEO, GEO and LLMSEO?

Very little in practice. AEO (answer engine optimisation) and GEO (generative engine optimisation) are used interchangeably for being cited in AI answers. LLMSEO or LLMO leans toward the model-knowledge layer. All three are facets of AI SEO. Do not let the acronyms distract you from the work.

Is AI SEO replacing traditional SEO?

No. AI SEO is the part of SEO that the new AI entry points make explicit. The same content quality wins both. The payoff just shifts from clicks to citations and brand presence in AI answers.

What is the single most important AI SEO move?

A consistent, machine-readable entity plus answer-first extractable content. AI assembles answers from clear, repeated facts about who you are and what you claim. Make those facts trivial to find and verify everywhere a machine looks.

How do I measure AI SEO?

Measure grounding, not clicks. Track the AI Crawler to AI Assistant ratio per URL in your server logs (the indexed-to-cited signal) and your Bing AI citation share. AI referral traffic is real but tiny, so it lags by weeks.

Does the AI SEO cheatsheet actually get results?

The tests behind it are published on this site rather than claimed here: the AI instructions page test that ChatGPT cited within 48 hours, and the 100-brand data set in the ChatGPT traffic analysis. Treat the cheatsheet as the checklist those tests produced, not a guarantee. Run it on one page, measure citations for 30 days, then scale.