Traditional SEO tells you to chase keywords. Generative Engine Optimization (GEO) asks a different question: what makes an AI model trust your page enough to cite it in an answer?
After running structured experiments across ChatGPT, Perplexity, and Google AI Overviews for six months, I've narrowed it down to 11 concrete signals. None of them require expensive tools. Most can be applied to an existing page in under an hour.
Use this as a page-by-page checklist before publishing, and again when auditing pages that should be getting cited but aren't.
How AI Engines Decide What to Cite
Before the checklist, a quick mental model. AI engines don't "rank" pages the way Google does. They retrieve candidate pages based on semantic relevance, then decide which ones to cite based on a few key trust signals:
- Comprehensiveness — does the page cover the topic's full entity graph, or just the surface query?
- Extractability — can the model pull a clean, self-contained answer from a section without needing surrounding context?
- Authority signals — does the page demonstrate first-hand knowledge, cite sources, and carry structured metadata?
Every item on the checklist maps to one of those three factors. I've marked each with [C], [E], or [A] so you know which problem you're solving.
The 11-Point Checklist
1. Write a Direct Answer in the First 100 Words [E]
AI models are extracting answers, not summarising full articles. If a page doesn't contain a clean, direct answer to the target query within the first 100 words of body content, it rarely gets cited — even if the rest of the article is excellent.
Action: After your intro paragraph, write one sentence that directly answers the primary question. Treat it like a definition. Example: instead of "GEO is a growing field…", write "GEO (Generative Engine Optimization) is the practice of optimizing content to appear as a cited source in AI-generated answers."
2. Use Explicit H2/H3 Structure That Mirrors Search Questions [E]
AI retrieval chunks documents by heading. A heading like "How long does GEO take to show results?" is far more likely to surface in a ChatGPT answer to that exact question than a heading like "Timeline considerations."
Action: Pull the "People Also Ask" questions from Google for your topic. Make each a verbatim H3. Frase and Surfer both surface these automatically.
3. Cover Every Major Related Entity [C]
AI models have an internal map of how concepts relate — what researchers call a knowledge graph. A page about GEO tools that doesn't mention "structured data", "entity disambiguation", or "citation rate" is missing nodes that the model expects to see on a comprehensive resource.
Action: Use Surfer or Clearscope's NLP entity panels to identify which related terms and concepts appear in top-cited pages. Add any you're missing with at least one substantive mention (not just a keyword drop).
4. Add a FAQPage Schema Section [E + A]
FAQPage JSON-LD is the single highest-leverage schema type for GEO. Google AI Overviews pull FAQ answers directly. Perplexity's retrieval pipeline assigns extra weight to structured Q&A content. And because FAQ sections naturally produce self-contained extractable answers, they help on ChatGPT too.
Action: Add a <section class="faq"><dl>
block to every article. This site's schemaBuilder module auto-generates
the JSON-LD from it — no manual schema writing required.
5. Implement Article + BreadcrumbList JSON-LD [A]
Article schema tells crawlers your page's headline, author, publish date, and modification date — all trust signals. BreadcrumbList gives the AI model structural context about where the page sits in the site hierarchy, which correlates with topic authority.
Action: Every article page needs both. Check the generated source
of your pages and confirm application/ld+json scripts for both types
are present in the <head>.
6. Show an Explicit Author With a Named Byline [A]
EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) matters to AI citation models just as it does to Google's quality raters. A named byline, an author bio that establishes relevant experience, and consistency across pages all contribute to the authority score.
Action: Every article needs a visible byline and an author bio
section or link. Your site.config.js authorName and
authorBio fields populate these automatically.
7. Include a Publish Date and a "Last Updated" Date [A]
Perplexity in particular filters for recency. Pages with no visible date are frequently skipped in favour of dated content, especially on topics that change quickly. Last-updated dates signal active maintenance — which AI models interpret as higher reliability.
Action: Add both dates to your article header and mirror them in
the datePublished / dateModified fields of your Article
schema.
8. Use Numbered Lists and Tables for Comparable Data [E]
When AI models encounter numbered lists and comparison tables, they parse them as structured data — making it far easier to extract a clean answer like "the top 5 GEO tools are…" rather than paraphrasing prose. Tables with clear column headers are especially powerful for Perplexity, which frequently surfaces tabular answers.
Action: Whenever you're comparing options, ranking items, or listing steps, use an ordered list or a properly-headed table. Never bury comparable facts in paragraphs.
9. Cite External Sources in the Body Text [A]
This one surprises people: citing external sources increases your citation rate. AI models treat outbound links to reputable sources (research papers, official documentation, established publications) as an authority signal — the same way academic papers gain credibility by referencing prior work.
Action: For every factual claim, link to a primary source. Target 3–5 outbound links per 1,000 words. Avoid linking only to your own content.
10. Achieve a Reading Level Accessible to a Non-Expert [E]
AI models are generating answers for a general audience. Pages written in dense jargon without plain-English explanations are less likely to produce quotes that fit naturally into a conversational answer. Aim for a Flesch Reading Ease score above 50 for most GEO topics.
Action: Read your page aloud. If a sentence would sound awkward spoken in a voice assistant answer, rewrite it. Hemingway App is a free tool that flags complexity issues inline.
11. Ensure Fast, Clean HTML — No JavaScript-Only Content [E]
AI crawlers vary widely in JavaScript execution support. Perplexity's crawler frequently skips JS-rendered content. If your key page sections (comparison tables, FAQs, key stats) only exist after JavaScript runs, they may never be indexed for citation purposes.
Action: Every critical piece of content must be in the raw HTML. JavaScript can enhance (sorting, filtering, interactivity) but must never be the only way content exists on the page.
Quick-Reference Checklist
Copy this into your publishing workflow:
| # | Signal | Factor | Time to implement |
|---|---|---|---|
| 1 | Direct answer in first 100 words | Extractability | 5 min |
| 2 | Question-phrased H2/H3 headings | Extractability | 20 min |
| 3 | Full entity coverage (NLP tool audit) | Comprehensiveness | 45 min |
| 4 | FAQPage schema section | Extractability + Authority | 15 min |
| 5 | Article + BreadcrumbList JSON-LD | Authority | 10 min |
| 6 | Named author with relevant bio | Authority | 5 min |
| 7 | Publish date + last-updated date | Authority | 2 min |
| 8 | Numbered lists and comparison tables | Extractability | 30 min |
| 9 | 3–5 outbound citations to primary sources | Authority | 15 min |
| 10 | Reading level: Flesch ≥ 50 | Extractability | 30 min |
| 11 | Key content in raw HTML (not JS-only) | Extractability | Varies |
Which Signals Matter Most?
If you're auditing existing content and want to triage, prioritise in this order based on my test results:
- FAQPage schema — highest single-signal citation lift across all three platforms
- Direct answer in the first 100 words — fastest to fix, immediate impact on Perplexity
- Entity coverage — biggest impact on ChatGPT citation rate (it's the most entity-aware of the three)
- Dates — critical for Perplexity recency filter; pages with no date are systematically penalised
Recommended Tools for This Checklist
You don't need all of these — pick one content optimizer and use it consistently:
- Entity coverage (signal 3): Surfer SEO or Clearscope
- Question headings (signal 2): Frase (surfaces PAA questions automatically)
- Reading level (signal 10): Hemingway App (free)
- Schema validation: Schema.org Validator (free)
For a full breakdown of how each tool scored in my GEO testing, see the Best GEO Tools for 2026 review.
Frequently Asked Questions
- Does this checklist apply to all AI engines, or just one?
- All 11 signals improve citation rates across ChatGPT, Perplexity, and Google AI Overviews, though with different weights. FAQPage schema is most impactful for Google AIO. Entity coverage matters most for ChatGPT. Dates and recency signals are most important for Perplexity. Applying all 11 covers you across the board.
- How do I know if my page is being cited in AI answers?
- The most reliable method is manual spot-checking: search your target queries directly in ChatGPT (with web browsing enabled), Perplexity, and Google. Look for your domain in the source citations. For Perplexity, you can use their API to run queries programmatically and log citation appearances over time.
- Can I apply this checklist to old content, or only new articles?
- Both. Existing pages often see the fastest gains because they may already have backlinks and index history — they just need the GEO signals layered in. Start by auditing your five highest-traffic pages against the checklist before writing new content.
- Will GEO optimization hurt my traditional Google rankings?
- In practice, no — the signals overlap significantly. FAQ schema, clear structure, entity coverage, and outbound citations are all positive SEO signals too. The one area to watch is readability: occasionally making content more conversational for GEO slightly reduces keyword density, but this is almost never a net negative for Google rankings.
- How often should I re-audit pages against this checklist?
- Quarterly for your top 10 pages by traffic, and before publishing any new article. The signals themselves are stable, but the competitive landscape shifts — a page that covered all entities six months ago may have gaps now if the topic has evolved.