Skip to content
AI Search Measurement

LLM Visibility Tools: What They Measure, and What They Can Only Estimate

How LLM visibility and GEO tools work, which metrics are observed versus modelled, how to evaluate one, and how to combine them with first-party data.

Analyst comparing a sampled AI answer log against a web analytics report

Field note

Illustrative image, AI-generated. It is not a client, dashboard or result. How we use imagery.

By XenGrowth EditorialPublished Reviewed 11 min read

Share this

Key takeaways

  • Most LLM visibility tools sample prompts and record answers; the result is an estimate that depends on the prompt set.
  • Ask any vendor which prompts, models, locations and repetitions produce each number.
  • Observed first-party data — referrals, crawler logs, Search Console and Bing reports — anchors the estimates.
  • Use tools to find gaps and track direction, not to promise a citation rate.

01

What does an LLM visibility tool actually measure?

Most LLM visibility tools run a set of prompts through AI assistants and search features on a schedule, then record whether your brand or pages are mentioned or cited and how often. The resulting share-of-voice or visibility score is a sample, not a census: it depends entirely on which prompts were chosen, which models were queried, from where, and how many times.

That does not make the tools useless. It means the number is only as good as the prompt set, and two tools can report very different scores for the same brand because they ask different questions.

What does an LLM visibility tool actually measure? comparison table
MetricHow it is producedTreat it as
Mentions / citationsSampled prompts, logged answersEstimate tied to that prompt set
Share of voiceYour mentions ÷ all brands in sampleRelative, directional signal
SentimentModel classifies answer textRough indicator, check samples
AI referral visitsYour analyticsObserved, but under-attributed
Crawler requestsYour server logsObserved access, not citation

02

How do you evaluate a GEO or LLM visibility tool?

Ask for the methodology in writing: how prompts are selected, whether you can supply your own, which platforms and model versions are covered, what locations and languages, how often each prompt is repeated and how answer variability is handled. Ask to see raw answers behind any score.

Check that the tool separates brand mentions from linked citations, and that it lets you export data. A dashboard you cannot audit is hard to defend when a number moves.

  • Prompt set visible and editable
  • Platforms, versions and locations stated
  • Repetitions per prompt and variance shown
  • Raw answers available behind every score
  • Export for your own analysis

03

Which first-party data should sit alongside the tool?

Your own data shows what actually reached you. Analytics can capture visits referred by AI assistants, though many arrive without a referrer. Server logs show which AI crawlers fetched which pages; OpenAI and Perplexity document their crawler user agents. Google includes AI features in Search Console performance data, and Bing Webmaster Tools reports AI performance for Copilot and Bing.

Read these together. A tool may show rising mentions while referrals stay flat, which suggests visibility without clicks; logs may show a crawler never reaching the pages you most want cited.

04

What should you expect from GEO work?

Be cautious about promised uplifts. The academic research behind GEO measured how much of an answer a source earned once it was already retrieved, and a later benchmark found most tested optimisation tricks produced no significant gain. Visibility depends on being crawlable, indexed and genuinely useful first.

Use tools to find prompts where competitors appear and you do not, then fix the underlying pages and third-party presence. Track the direction over months rather than reacting to weekly swings.

Was this useful? Send it to someone who owns this problem.

Primary sources and further reading

Use the source material to validate details against your own context and current platform configuration.

This field note follows the XenGrowth editorial policy: primary sources where available, visible limitations, material review dates, and no invented first-hand experience.

Stay with the problem

Explore AI search & GEO