Trust AgencyTrust Agency
Technical & AI search

GEO visibility scorecard

Generative engines do not rank pages, they assemble answers and cite sources. Being cited depends on four things: whether a crawler can reach you, whether a model can extract a self-contained answer, whether the claim is verifiable, and whether the wider web describes you consistently. This scorecard weights all four and tells you which gap to close first.

GEO visibility scorecard

Tick what is genuinely true of your site today. The score weights each item by how much it influences whether an answer engine can find, parse, trust and quote you — and the gap list tells you what to fix first.

Access & crawlability

GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended. Blocking them is a choice, not a default.

Most answer-engine crawlers do not execute JavaScript. If the text is not in the source, it does not exist for them.

Not an official standard, cheap to publish, and it removes ambiguity about what the site is.

A crawler that cannot read the answer will quote the competitor who published it openly.

Structure & extractability

Structured data is the cheapest way to state facts unambiguously.

Extractable answers get quoted; narrative introductions get skipped.

A section that only makes sense after reading the previous one cannot be lifted as a citation.

Tabular facts survive extraction far better than the same facts in prose.

Evidence & expertise

The single strongest predictor of being cited: being the source, not a summary of sources.

Quantified statements are quoted disproportionately often in generated answers.

Author identity is part of how both Google and answer engines weigh trust.

Answer engines favour recent sources on anything that changes.

Off-site authority

Wikipedia, Wikidata, industry directories, trade press, G2-style review sites, Reddit threads.

Contradictory descriptions split the entity and make the model hedge.

Generated recommendations lean heavily on off-site consensus.

You cannot improve a visibility you never measure.

GEO readiness score

0 / 100

Tier

Largely invisible

Criteria met

0 / 16

Score by area

Access & crawlability0 %
Structure & extractability0 %
Evidence & expertise0 %
Off-site authority0 %

Fix these first

  1. AI crawlers are explicitly allowed in robots.txt
  2. Main content is in the server-rendered HTML, not injected by JavaScript
  3. The site publishes original data, benchmarks or research nobody else has
  4. Pages carry accurate schema.org markup (Organization, Article, FAQ, Product)
  5. Pages answer the question in the first 100 words, before the build-up

No search engine has published a GEO ranking factor, and anyone who claims otherwise is selling something. This scorecard weighs what repeated citation analyses keep finding: machine-readable pages, self-contained answers, verifiable evidence and a consistent off-site identity. Those are also plain good SEO, which is the point — GEO is not a separate discipline, it is the same discipline judged by a stricter reader.

How to improve visibility in AI answers

  1. Check access first — Confirm your robots.txt allows the answer-engine crawlers you want, and that the content is in the server-rendered HTML.
  2. Make answers extractable — Answer the question in the first paragraph, use question headings, and put comparisons in real tables.
  3. Publish something only you have — Original data, benchmarks and named expertise are the strongest predictors of being quoted.
  4. Align the off-site story — Directories, review sites, trade press and encyclopaedias should describe the same entity in the same words.
  5. Measure citations monthly — Track your core questions across the major assistants and record who gets cited instead of you.

Why the four groups are weighted the way they are

Access comes first because everything else is theoretical without it. Most answer-engine crawlers do not execute JavaScript, so content injected client-side simply does not exist for them; a site behind a consent wall or an aggressive bot filter is invisible no matter how good the writing is. Blocking these crawlers is a legitimate business decision — but it should be a decision, not something a security vendor enabled by default.

Extractability comes next because generated answers are assembled from passages, not pages. A section that begins with a direct answer, sits under a heading phrased the way a person asks the question, and makes sense lifted out of context is a section that can be quoted. Long narrative build-ups, comparisons buried in prose and facts that depend on the previous paragraph are all invisible to that process.

Evidence and off-site authority decide whether the passage is trusted once found. Models weigh corroboration heavily: a claim repeated across independent sources, attributed to a named expert and dated, survives where an anonymous assertion does not. This is why original research outperforms summaries so decisively, and why the same directory listings, review profiles and trade coverage that support classic SEO also support citation.

What is honestly unknown

No answer engine publishes its citation criteria, and the vendors selling GEO certainty are inferring from the same public analyses everyone else reads. What those analyses agree on is directional: cited sources skew towards pages that are accessible, structured, specific, recent and corroborated elsewhere. What they disagree on is magnitude, and any scorecard — including this one — is a weighting of judgement, not a measurement.

The reassuring part is that the checklist contains almost nothing you would not do anyway. Server-rendered content, accurate structured data, clear headings, real evidence and consistent entity data are the same recommendations that have applied to search for a decade. GEO is less a new discipline than the same discipline judged by a reader with no patience for preamble.

The one genuinely new habit is measurement. Nothing in your analytics tells you that an assistant summarised your page without sending a visit. Ask your core questions across the major assistants once a month, record who is cited, and treat competitors who appear instead of you as the gap analysis. Our guide on getting cited by AI sets out a repeatable process, and the GEO tools category lists the products that automate the tracking.

  • Decide deliberately whether to allow AI crawlers — do not inherit a default
  • Put the answer in the first 100 words of every page that answers something
  • Use real tables for comparisons and specifications
  • Publish at least one dataset or benchmark only you can produce
  • Keep the entity description identical across every off-site profile
  • Track citations monthly; analytics will never show them to you

Frequently asked questions

What is GEO, and is it different from SEO?

Generative engine optimisation is the practice of being found and quoted inside AI-generated answers rather than ranked in a list of links. The techniques overlap heavily with SEO — crawlability, structure, evidence, entity consistency — but the objective differs: a citation with no click still builds awareness, and a page that ranks first but cannot be extracted may never be quoted.

Should I block AI crawlers?

It depends on your business model. A publisher whose revenue comes from page views has a real argument for blocking; a services company that wants to be recommended has almost none. What matters is that the choice is deliberate and revisited, since a blanket block also removes you from the answers your prospects are reading.

Does llms.txt actually do anything?

It is not an official standard and no engine has committed to reading it, which is why it carries a small weight here. It costs almost nothing to publish, states plainly what your site is and where the canonical entry points are, and removes ambiguity for anything that does read it. Treat it as cheap insurance, not as a lever.

Other calculators

SEO health self-audit

SEO health score

Core Web Vitals ROI calculator

Core Web Vitals

SEO traffic forecast calculator

Traffic forecast

Go deeper