The AI-Visibility Standard
Nobody publishes an official standard for how a business gets found by AI. So we built a transparent one — from the actual ratified specs and peer-reviewed evidence — and we keep it current in the open. Every factor we score, and the real standard behind it, is on this page.
Updated July 8, 2026 Maintained by Fusion Tech AI
The honest truth, up top
There is no official AI-visibility standard.
No standards body — not the IETF, W3C, ISO or WHATWG — and no AI vendor — not OpenAI, Anthropic, Google or Perplexity — publishes a canonical AI-visibility score or an official rubric to compute one. Google says the opposite of a hidden checklist:
“There are no additional requirements to appear in AI Overviews or AI Mode… You don’t need to create new machine readable files, AI text files, or markup… There’s also no special schema.org structured data that you need to add.” — Google Search Central, AI Features and Your Website, updated 2025-12-10
So anyone selling you “the official AI score” is bluffing. What does exist is a stack of real, published standards that decide whether AI and search engines can even read and trust your site — and a small body of peer-reviewed evidence on what earns a citation. This document measures your site against those actual standards and rolls them into one number, and tells you flat-out which parts are ratified standards, which are industry conventions, and which are unproven experiments. That transparency is the authority. Every AI-visibility tool on the market is a proprietary composite exactly like this one; the difference is we show you the ingredients.
The standard · section 1
Every factor we score — and the real standard behind it
This is the centerpiece. These are the exact factors our free scanner measures, with the primary standard, its owner, its date, our weight, and a plain-English reason. Weights are ours — a proprietary composite, because no official one exists — and they sum to 100 across the point-earning factors. Two factors are deliberately not additive: crawl access is a pass/fail gate, and page performance is measured for information but scored at zero.
| Factor | Weight | Real standard | Owner & date | Why it matters |
|---|---|---|---|---|
| Evidence density statistics, cited sources, quotations On-page | 36 | Generative Engine Optimization (GEO) | Aggarwal et al. — Princeton / Georgia Tech / Allen Institute for AI / IIT Delhi arXiv:2311.09735 · Nov 2023 · KDD 2024 | The peer-reviewed study found adding statistics, citations and quotations lifted a page's visibility in generative answers by up to ~40%. This is the single strongest lever on your own page. Results vary. |
| Answer formatting FAQ / Q&A schema, question headings, lists On-page | 24 | Schema.org FAQPage + extractability | Schema.org steering group (Google / Microsoft / Yahoo / Yandex) Schema.org v30.0 · 2026-03-19 | Answer-shaped content is easier for an engine to lift a direct answer from. Profound's data indicates Claude favors structured, bulleted pages. |
| Structured data Schema.org JSON-LD On-page | 12 | Schema.org vocabulary | Schema.org (Google / Microsoft / Yahoo / Yandex) v30.0 · 2026-03-19 | Helps engines understand your content and become eligible for rich results. Google states no special schema is required for AI answers — so we treat it as comprehension hygiene, not a proven citation boost. |
| Freshness published / updated dates On-page | 12 | Recency signals (documented, magnitude directional) | Perplexity + Google (own documentation) ongoing | Perplexity and Google's AI Overviews favor recent content. A visible, honest updated date helps; the exact size of the effect is directional, not proven. |
| Structured contact name / address / phone — local businesses only On-page | 12 | Schema.org LocalBusiness + Google local eligibility | Schema.org / Google Search v30.0 · living doc | For 'near me' questions, clear name-address-phone helps an engine anchor your business as an entity. Excluded from scoring (weights renormalize) when the site is not a local business. |
| Open Graph / Twitter social + link-unfurl meta On-page | 3 | Open Graph Protocol | Meta (ogp.me) v0.9 · since 2010 | Controls how your link renders as a rich card when it is shared or cited. A de-facto standard, never ratified by a formal standards body. |
| llms.txt AI-summary file Emerging | 1 | llms.txt proposal (emerging, unproven) | Jeremy Howard / Answer.AI proposal · 2024-09-03 | An emerging convention. Ahrefs found 97% of llms.txt files received zero requests, and Google confirms it does not use them. We keep it at 1 point as cheap future-proofing and never oversell it. |
| Crawl access robots.txt — AI inference bots Gate | GATE | Robots Exclusion Protocol | IETF RFC 9309 · Sept 2022 | The one fully ratified standard in the stack. If robots.txt blocks GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the like, an engine literally cannot read or cite you. So this is a pass/fail gate that multiplies the score down — never points you earn for merely allowing bots. |
| Page performance Lighthouse signals / Core Web Vitals Info only | 0 | Lighthouse performance scoring | Google (Chrome) TBT 30 / LCP 25 / CLS 25 / FCP 10 / SI 10 | The closest thing to an official website score, and a real search-ranking factor — but not an AI-citation lever. We measure and display it, and deliberately score it at zero so a fast, empty page cannot look AI-ready. |
Point-earning weights sum to 100 when a site is local (evidence density 36 · answer formatting 24 · structured data 12 · freshness 12 · structured contact 12 · Open Graph 3 · llms.txt 1). For non-local sites, structured contact is excluded and the remaining weights renormalize. Crawl access applies a penalty multiplier when AI inference bots are blocked; page performance is never scored. Recalibrated 2026-07-08 against the Princeton GEO paper and the Ahrefs, Semrush and Profound studies.
The standard · section 2
What actually drives an AI citation
Here is the uncomfortable, evidence-based split. The biggest driver of whether AI cites you is off-page — and a scanner cannot see or fix it. The biggest thing you actually control is on-page content substance. We rank both honestly.
“Strategies with higher impact are those that increase the intrinsic quality and credibility of the content, not those based on mechanical repetition or lexical over-optimization.” — Aggarwal et al., GEO: Generative Engine Optimization, Princeton / GA Tech / AI2 / IIT (2023)
On-page — what you (and we) can fix
Proven to move visibility 15–40% for the pages that get retrieved (Princeton GEO). Results vary.
- 1 · Content substance
Statistics, cited sources and direct quotations. The Princeton GEO study's top methods — Cite Sources, Quotation Addition, Statistics Addition — each achieved 30–40% relative improvement in generative-answer visibility; the “+41%” figure for adding statistics is the most-repeated single number (directional).
- 2 · Extractability & structure
FAQ / Q&A blocks, question-style headings and lists. Profound's citation data indicates Claude is roughly 30% more likely to cite structured, bulleted pages.
- 3 · Freshness
Recent, honestly-dated content. Perplexity and AI Overviews weight recency; the size of the effect is directional, not proven.
- 4 · Crawlability & schema (hygiene)
Necessary conditions, not levers. Being reachable and machine-readable earns you eligibility — the studies credit it with roughly zero citation lift on its own.
Off-page — authority a scan can't fake
The larger driver, per the biggest public datasets — and invisible to any on-page scanner.
- YouTube mentions — the #1 correlate
Ahrefs' 75,000-brand study put YouTube mentions at a 0.737 correlation with AI visibility — the strongest single factor measured.
- Brand mentions ≈ 3× backlinks
Third-party branded web mentions correlated ~0.664 versus ~0.19–0.24 for raw backlinks. A backlink tells AI where to go; a brand mention tells it what to trust. (Correlation, not causation.)
- The citation pool leans on communities
Semrush's June 2025 study of ~150,000 AI citations found Reddit at 40.1%, Wikipedia at 26.3% and YouTube at 23.5% of references — all off-page to a small-business site.
We give AI engines every on-page signal they need — clear, substantive, extractable, crawlable content. What each engine does with that, and how it weighs your off-site reputation, is up to them, and results vary. A scan cannot measure or fix off-page authority; that is a separate, longer game.
The standard · section 3
Do it yourself — the honest, free version
We give away the truth. If you are technical, here is genuinely how to move the on-page factors above. None of this requires us; it only requires the work.
How do I add statistics and citations that AI will use?
Replace vague claims with specific numbers (“trusted by many” → “serving 1,200 customers since 2019”), and link out to the primary source for each factual claim — a study, a government page, a manufacturer spec. Add a short expert quotation where you have one. These three moves are the Princeton GEO study's highest-impact methods.
How do I make my page answer-shaped?
Write real questions as your H2 and H3 headings (the ones customers actually ask), answer each in the first sentence beneath, and use short lists for anything an assistant might read aloud. Add FAQPage schema so the Q&A structure is machine-explicit. This page is built exactly that way.
How do I add the right structured data?
Add Schema.org JSON-LD for your Organization or LocalBusiness, your Services, and a FAQPage. Validate it in Google's Rich Results Test. Remember Google's own position: schema is a comprehension aid and eligibility mechanism — it will not, by itself, get you cited.
How do I keep it fresh and crawlable?
Put a visible, honest “Updated <date>” on pages you actually revise (never fake it), and confirm your robots.txt does not block GPTBot, ClaudeBot, PerplexityBot, Google-Extended or Bingbot — because if it does, none of the above can be read. Check your robots.txt against RFC 9309.
Want to know your site’s score?
Run the free scanner. It measures your site against every on-page factor in the table above and shows you exactly what it found — your own stats, your cited sources, your schema — with no jargon and no lock-in. Or, if you would rather we do the work, we rebuild the whole engine.
Website Recycling rebuilds start at $997 one-time plus $50/month maintenance. 60-day money-back, no questions asked. You keep your content, brand, domain and data. Results vary.
Questions, answered plainly
Frequently asked questions
Is there an official AI-visibility standard?
What actually gets a page cited by AI?
Does adding an llms.txt file get me cited by AI?
Does schema markup boost my AI ranking?
Can a website scanner measure everything that affects AI visibility?
How often is this standard updated?
Version history
Changelog
A living standard. Every change to the factor table is logged here with its date, so you can see exactly what moved and when.
Initial publication. Factor table established from RFC 9309, Schema.org v30.0, Open Graph v0.9, and the Princeton GEO paper. Weights recalibrated so content substance dominates (evidence density 36, answer formatting 24), crawl access became a pass/fail gate rather than free points, and page performance was set to score zero as an AI-citation factor.
References
Sources
Every claim on this page traces to a primary source. Here they are, with dates.
- RFC 9309 — Robots Exclusion Protocol · IETF · Sept 2022 · rfc-editor.org/info/rfc9309
- Schema.org releases (v30.0) · Schema.org steering group · 2026-03-19 · schema.org/docs/releases.html
- The Open Graph Protocol (v0.9) · Meta · since 2010 · ogp.me
- Sitemaps protocol (v0.9) · sitemaps.org working group · Nov 2006 · sitemaps.org/protocol.html
- Lighthouse performance scoring · Google (Chrome) · TBT 30 / LCP 25 / CLS 25 / FCP 10 / SI 10 · developer.chrome.com/docs/lighthouse/performance/performance-scoring
- Aggarwal et al., “GEO: Generative Engine Optimization” · Princeton / GA Tech / AI2 / IIT · arXiv:2311.09735 · Nov 2023 (KDD 2024) · arxiv.org/abs/2311.09735
- Ahrefs, “Top Brand Visibility Factors in ChatGPT, AI Mode & AI Overviews (75k Brands)” · Dec 12, 2025 · ahrefs.com/blog/ai-brand-visibility-correlations
- Ahrefs, llms.txt study (137,210 domains; 97% zero requests) · 2026 · ahrefs.com/blog/llmstxt-study
- Semrush AI citation study (~150,000 citations; Reddit 40.1% / Wikipedia 26.3% / YouTube 23.5%) · June 2025 · semrush.com/blog/the-ghost-citations-study
- Profound, “AI Platform Citation Patterns” (680M+ citations) · Aug 2024 – Oct 2025 · tryprofound.com/blog/ai-platform-citation-patterns
- Google Search Central, “AI Features and Your Website” · updated 2025-12-10 · developers.google.com/search/docs/appearance/ai-features
- OpenAI, crawler / robots.txt controls (GPTBot, OAI-SearchBot, ChatGPT-User) · developers.openai.com/api/docs/bots
- llms.txt proposal · Jeremy Howard / Answer.AI · 2024-09-03 · llmstxt.org
Note on evidence: the Ahrefs correlations are brand-level associations, not causation. The granular per-method GEO percentages that circulate in secondary write-ups (e.g. “+41% statistics”) are directional interpretations; the Princeton paper's own headline figures are up to ~40% overall and 30–40% for its top methods. The “content under 30 days = 3.2× citations” figure is widely repeated but not traced to a primary dataset. We label these plainly rather than present them as certainties. Results vary.
Maintained by Fusion Tech AI · applied by Website Recycling · v1.0 · Updated July 8, 2026.