FAQ

How do you measure a brand's visibility in AI answers?

Four metrics. (1) Mention rate: fix a set of questions your buyers actually ask (about 30, bilingual), sample the major engines weekly, and count in how many answers the brand appears. (2) Citation mix: of the sources cited, how many are your own properties versus third parties. (3) Sentiment and accuracy: whether mentions recommend, neutrally describe or criticise you — and whether the facts are right. (4) AI-referred conversion: visits and inquiries arriving via AI channels, attributed separately. One caution: single-day data is noise — citation rates fluctuate up to 35% within a day, so run a full quarter of baseline before reading trends. Negency's stats backend tables bot crawling and AI-referred leads separately precisely to ground this measurement.

Figure 01 · Five technical-visibility dimension means

  1. 01

    Crawl reachability

    Bots can reach and fetch the site

    91.1/100
  2. 02

    Performance

    Response and rendering basics

    87.3/100
  3. 03

    Trust signals

    Contact, company identity and verifiable facts

    82.1/100
  4. 04

    Entity structure

    Facts structured so the entity is recognisable

    71.1/100
  5. 05

    Machine-readable surface

    Structured data and text machines can read directly

    69.2/100 Weakest dimension · median 68.5
Dimension means from the Negency Flagship Report 2026 (Issue 1), measured across 30 Chinese exporter websites in a single-day sweep on 30 October 2026. These are technical-visibility scores, not brand mention rates — mention rates need separate fixed question-set sampling. Machine-readable mean 69.2, median 68.5 — two different statistics.

Mention rate: a fixed question set, sampled repeatedly

Questions must come from real buyer behaviour, not brand-name searches. Once fixed, the set never changes (changing it breaks the baseline); sample weekly on identical terms, per engine. Expect a low baseline — a brand with zero AI-channel presence starts near zero mentions.

Clicks are shrinking. Pew Research Center (July 2025) followed 900 US adults across 68,879 Google searches in March 2025: when an AI summary appeared, users clicked a traditional result on 8% of visits, against 15% without one; links inside the summary were clicked on just 1% of visits.

The arithmetic (illustrative numbers, not measured data): a recommendation mention in 6 of 40 questions is a 15% mention rate. Recommendation-style mentions — the answer actively listing you as a supplier option — are the headline GEO metric.

Citation mix: owned vs third-party

If every citation points to your own pages, the off-site ecosystem has not started; if citations point to stale or wrong third-party pages, fix facts before promotion.

Ahrefs' study of 75,000 brands found branded web mentions correlated with AI visibility at 0.66–0.71, far above backlink counts and page counts (about 0.19–0.22). A mix with almost no third-party share means your visibility rests entirely on your own pages.

Accuracy before sentiment

A positive mention with wrong facts is worse than no mention: errors get re-cited by other systems. Keep a dedicated column for factual errors and run the correction path when one appears.

A typical case: an AI answer states that a company "offers free samples" when its actual policy charges for them. Buyers arrive quoting the wrong policy, and other AI systems repeat the error as fact. That is what the factual accuracy rate is for: of the answers that mention you, how many get the key facts right.

Conversion attribution: visibility must land as inquiries

Log AI-referred visits separately (referrer domains, agent identifiers) and preserve source parameters into the lead form. At quarterly review, plot the visibility curve next to the inquiry curve.

Track attribution on two levels: visits (AI-referred volume, reconciled between logs and analytics) and inquiries (forms and consultations carrying an AI source tag).

A reproducible measurement method, in five steps

  • Fix the question set: 20–60 real questions across the customer journey (category, comparison, scenario, brand), mixed by market language, then frozen — change the set and the data stops being comparable;
  • Fix the sampling rules: a set engine list, sign-in state, region and language settings; sample every question on every engine at least 3 times and take the majority — a single result is noise;
  • Record four raw facts: whether the brand is mentioned (recommendation-style mentions flagged separately), whether your pages are cited, whether the description is accurate, and how competitors do;
  • Compute five metrics: recommendation mention rate, citation rate, share of citations, factual accuracy rate, and AI-referred visits;
  • Baseline before trend: run four full weeks of baseline, then re-test weekly; never conclude from a single week — read the four-week moving average.

A technical score is not a mention rate

Do not mix the two kinds of score. The Negency Flagship Report 2026 (Issue 1) measures technical visibility — whether AI can reach and read a website. Across 30 Chinese exporter websites (sampled 30 October 2026), the mean total was 80.1/100 and the median 80 (range 55–98), with 7 sites above 90 and 6 below 70. (Sampling note: the sample is the global homepages of 30 Chinese exporter websites, measured in a single-day sweep on 30 October 2026 with the Negency audit engine — about 30 signals in five groups. The report measures technical visibility only — whether AI can reach and read the site — not brand mention rates in AI answers.) A high technical score only proves machines can read you — it does not mean AI answers will mention you. The technical score is the entry ticket; the mention rate is the outcome.

The three most common measurement mistakes

  • Testing only brand questions: customers ask category and scenario questions; brand-question mention rates are inflated by construction;
  • Treating one screenshot as evidence: the same question at a different time or account can return a different answer — sample repeatedly;
  • Counting mentions but not accuracy: being mentioned wrongly (pricing, positioning, someone else's case) hurts more than not being mentioned — track the accuracy rate separately.

Run a free AI visibility audit first for your technical baseline, then build your question set and measure mention rate with the five steps above.

Related questions

How is GEO different from SEO?

SEO optimises a page's rank in a list of links, aiming for clicks. GEO (Generative Engine Optimization) optimises how discoverable, understandable, trustworthy and citable a company is inside AI-generated answers — the goal is to be in the answer itself. They share technical foundations (crawlability, structured data, content quality), but GEO additionally requires governed entity facts, a machine-readable surface (llms.txt, Markdown mirrors) and a third-party mention ecosystem. A site that ranks well is not automatically visible to AI: engines cite the companies whose facts they can fetch, parse and verify. The concrete differences sit in four dimensions — goal, object, content structure and evaluation — compared item by item in the body below.

How can Chinese companies going global build AI search visibility?

For Chinese companies going global, the visibility gap is rarely content volume — it is the machine-readable surface. Our audit of 30 Chinese outbound companies found machine-readability the weakest category (69.2/100 average): 17 had no JSON-LD on the homepage, 15 had a missing or thin sitemap, 13 had a missing or duplicated H1. The path has four steps: audit bot reachability and machine-readability first; govern entity facts so Chinese and English statements agree; publish question-shaped content built on real overseas buyer questions; then earn third-party mentions (media, communities, partners) and verify with periodic resampling against a fixed question set.

What should a corporate website change for the AI search era?

Five changes. (1) Bot reachability: robots.txt explicitly allows AI crawlers and live fetches are not stopped by firewalls or challenge pages. (2) A machine-readable surface: an llms.txt fact file (optional aid), a complete sitemap, Markdown versions of core pages (optional). (3) Entity structure: JSON-LD naming entity types such as Organization, clean title/H1 discipline, hreflang pairing on bilingual sites. (4) Question-shaped content: pages organised around buyer questions, short answers first, FAQ schema attached. (5) Measurability: analytics that separate human and bot traffic and attribute AI-referred inquiries. Negency's own site was rebuilt to exactly this standard, and the public audit tool scores against it.

What is llms.txt, and should every company website have one?

llms.txt is a plain-text brief at the site root written for AI systems: who the company is, what it does, where the core pages live, and the canonical facts and definitions. It does not by itself raise citation rates — in Ahrefs' May 2026 study of 137,210 domains, 97% of published llms.txt files received zero requests that month, so its direct effect is near zero — but it is the cheapest fact-governance move available: one file anchors a single version of the truth that crawlers can rely on. Verdict: worth adding, useless alone; it works as part of entity-fact governance, structured data and question-shaped content. Negency's llms.txt is a full fact bible: company definition, offering boundaries, sourced cases and citation rules.

What should an enterprise prepare before deploying AI agents?

Four preparations, and we advise against starting without any of them. (1) Scenario selection: pilot on a workflow that is frequent, comparatively rule-clear and verifiable — not the most glamorous one. (2) Knowledge and data foundations: govern the knowledge base first (deduplication, versioning, permissions); a dirty knowledge base only scales errors. (3) Governance design: permission boundaries, human approval points and audit logs decided before launch, not after. (4) A measurement baseline: record the manual process's time and error rates before the pilot, or value can never be proven afterwards. Negency's method is a small controlled pilot, verified before scaling, with governance as a design input rather than a patch.