The 12-Point GEO Checklist: Is Your Page Ready to Be Cited by AI?

Most GEO advice is spread across long guides you read once and forget. This is the opposite: a checklist you can actually work through, point by point, against any page on your site — in under an hour.

For each item: what to check, why it matters, and how to tell if you pass.

Access: can AI engines even read your page?

1. AI crawlers aren't blocked in robots.txt

Open yourdomain.com/robots.txt and look for the search/retrieval bots that power AI citations: OAI-SearchBot, ChatGPT-User (OpenAI), Claude-SearchBot, Claude-User (Anthropic), PerplexityBot, Perplexity-User (Perplexity). If any of these have a Disallow: / under them, you're removed from that engine's answers.

Pass: none of the search/retrieval bots are disallowed. (Whether you block training crawlers like GPTBot or ClaudeBot is a separate business choice — it doesn't affect live citations.)

2. Your CDN or firewall isn't silently blocking bots

robots.txt is only half the story. CDN and security layers (Cloudflare especially) often block AI crawlers by default, above the level your robots.txt controls — and audits in 2026 found roughly 41% of B2B sites still blocking at least one major AI bot, usually without knowing.

Pass: your firewall/CDN bot settings explicitly allow known AI crawlers, or you've confirmed AI bot traffic appears in your server logs.

3. Your content exists in the raw HTML

Some AI crawlers read the raw HTML response, not the JavaScript-rendered page. View your page source (Ctrl+U) and check whether your actual content — headings, paragraphs, facts — is present, or whether the source is just an empty shell waiting for JavaScript.

Pass: your key content is visible in the page source, not only after rendering.

Structure: can the answer be extracted?

4. The direct answer appears in the first 40-60 words

AI engines favor pages that state their answer immediately. If your page takes three paragraphs to get to the point, an AI evaluating candidates for a citation has often already moved on to one that just said the thing.

Pass: a stranger reading only your first two sentences could correctly state what this page's answer is.

5. Headings are questions where natural, and properly nested

Headings that mirror real questions ("How much does carpet cleaning cost?") map directly onto the queries AI engines are answering. And clean hierarchy (H1 → H2 → H3, no skipped levels) helps machines parse which content belongs to which topic.

Pass: your headings would make sense as search queries, and the hierarchy is clean.

6. Paragraphs are short and self-contained

Walls of text extract poorly. 2-4 sentence paragraphs, each making one clear point, extract cleanly — and read better for humans anyway.

Pass: no paragraph on the page runs past 5-6 sentences.

7. There's a genuine FAQ or Q&A section

Question-and-answer formatting is one of the easiest structures for AI systems to lift into their answers. The questions need to be real ones your customers ask — not filler written to occupy space.

Pass: your page (or site) answers the actual questions customers ask you, in explicit Q&A format.

Trust: is there anything worth citing?

8. Fact density is high enough

The original Princeton research that defined GEO found that adding statistics, citations, and quotations were among the highest-impact improvements for AI visibility. A useful working target: a concrete fact, number, or named source roughly every 150-200 words.

Pass: count your page's concrete facts and divide by word count. Vague pages ("high quality, fast service") fail; specific pages ("$100-280 typical cost, 2-3 day scheduling") pass.

9. Claims are attributed to named, dated sources

"Studies show" gives an AI nothing to verify. "A 2026 CapstonAI audit found" gives it a checkable reference — and checkability is what separates citable pages from ignorable ones.

Pass: every significant claim on the page either comes from your own first-hand knowledge or names its source.

10. Schema markup tells machines what the page is

JSON-LD structured data removes ambiguity: Article (with author and date) for content, Product or SoftwareApplication for products, LocalBusiness for local services, FAQPage for Q&A sections. Google retired the visual FAQ rich results in May 2026, but the markup itself remains valid and machine-readable.

Pass: your page has schema matching its actual type, and it validates without errors.

Freshness: does it look alive?

11. The page shows a visible, recent date

AI engines weigh freshness, particularly for topics where currency matters — and cross-reference against newer sources. An undated page, or one last touched years ago, loses citations to fresher equivalents even when its content is still accurate.

Pass: the page displays a published or updated date, and the content has been reviewed within the past year.

12. Your facts are consistent across the web

AI engines corroborate. If your website says one thing and your Google Business Profile, directories, or social profiles say another — different hours, addresses, prices, claims — that inconsistency reads as unreliability.

Pass: your core business facts match everywhere they appear.

---

Scoring yourself honestly

Count your passes out of 12:

• 10-12: Genuinely well-positioned. Your remaining gains are in depth and freshness maintenance.
• 6-9: Typical of a decent site that's never specifically optimized for AI. The access and structure items (1-7) are usually the fastest wins.
• 0-5: Your content is likely invisible to AI engines regardless of its quality — start with items 1-3, because nothing else matters until crawlers can read you.

The faster version of this checklist

Working through all 12 points manually takes under an hour per page — completely doable. But if you'd rather have it done in about a minute, this checklist is essentially a hand-run version of what aireadypage automates: paste a URL, get scored against 18 criteria (these 12, plus deeper technical checks), and receive a prioritized fix list ranked by impact. The first analysis is free.

Either way — run the audit. Most pages fail more of these points than their owners expect, and most of the fixes take an afternoon, not a rebuild.