The essentials
- What they do: automatically put a set of questions to several AI assistants, record the brands cited and the sources used, and track the trend.
- Criteria: assistants covered, number of questions tracked, reading frequency, source detection, competitor tracking, language and local market.
- 2026 pricing: €30 to €150 a month for a small business, €300 to €1,500 for a multi-client agency.
- The shared limit: none of them has access to the assistants' internal data; all of them work by repeated questioning, and therefore by sampling.
Tracking visibility in generative AI answers has become a real need as assistants take a growing share of searches. A category of tools has appeared to meet it. This article explains what they measure, how to tell them apart, what they cost and when a manual method is enough. The measurement approach itself is described in tracking visibility on ChatGPT and Perplexity.
What these tools really measure
No vendor has access to the assistants' internal data. All of them work the same way: they automatically put a set of questions to ChatGPT, Gemini, Perplexity and sometimes Copilot at regular intervals, then analyse the answers to extract the brands cited, the order they appear in and the sources referenced. It is sampling, with the limits that implies: answers vary from one run to the next, per-user personalisation isn't reproducible, and real query volumes remain unknown. A tool promising "AI search volumes" is extrapolating; a tool measuring a citation frequency across a sample is doing what is technically possible.
The selection criteria
| Criterion | What to check | Why |
|---|---|---|
| Assistants covered | ChatGPT, Gemini, Perplexity at minimum; Copilot and Le Chat depending on your customers | Answers differ a great deal from one model to another |
| Local market | Questions in your language, localised answers | Many tools are calibrated for English |
| Number of questions | 20 to 100 depending on the plan | Determines how fine the tracking is |
| Reading frequency | Weekly or monthly | Daily is pointless, variability dominates |
| Repetitions per question | At least three | Without them the reading measures chance |
| Source detection | The pages cited, not just the brands | This is what makes the tracking actionable |
| Competitor tracking | Comparative share of voice | The gap counts more than the absolute value |
| Export and API | Into an existing dashboard | Avoids one more tool to open |
Price ranges in 2026
| Category | Monthly price | Who for |
|---|---|---|
| Specialist GEO tools | €30 to €150 | A small business tracking one brand and a few competitors |
| GEO modules in SEO suites | Included, or €20 to €80 extra | Companies already using a rank tracking tool |
| Multi-client platforms | €300 to €1,500 | Agencies tracking several brands |
| Bespoke solutions (the models' APIs) | The cost of the calls plus development | Organisations with technical skills |
| The manual method | €0, 15 to 30 minutes a quarter | Micro-businesses and small companies with lower stakes |
The market is moving fast: the specialist players that appeared in 2024 and 2025 are consolidating, and the established SEO suites are gradually adding these functions. Checking language coverage and how fresh the models queried are matters more than the vendor's name. The wider marketing tools landscape is in the best digital marketing tools in 2026.
The manual method, often sufficient
- Fix ten to twenty questions: shortlisting, comparison and expertise, written the way a customer would ask them.
- Choose three assistants: ChatGPT, Gemini, Perplexity, in a neutral session with no history.
- Repeat each question three times: to smooth out the variability of the answers.
- Note four things: brands cited, order, sources used, justification given.
- Consolidate in a spreadsheet: one row per question, one column per quarterly reading.
- Decide: every absence points to content to produce, every source cited points to a page that works.
This protocol takes about thirty minutes a quarter and produces most of what a €100-a-month tool supplies. It becomes insufficient beyond fifty or so questions, or for multi-client tracking. The competitive reading is covered in share of voice in AI.
What no tool replaces
- Analysing the sources cited: knowing that a competitor is cited is useless without reading the page that earns them the citation.
- Fixing the content: the tools measure, they don't write the missing answers.
- Entity consistency: inconsistencies between directories, listings and your site are fixed by hand.
- The link to commercial results: no tool connects a citation to a signed customer; that reconciliation happens in the CRM.
- The complementary metrics: AI referral traffic in GA4 and brand searches in Search Console, to be tracked alongside; see GEO KPIs.
How GreenRed helps
Rather than juggling several tools, GreenRed's GEO tracking brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
What do GEO tracking tools measure?
They automatically put a set of questions to several AI assistants at regular intervals, then record the brands cited, their order and the sources referenced. None has access to the models' internal data: all of them work by sampling, with the limits that implies.
How much does an AI visibility tracking tool cost?
From €30 to €150 a month for a small business tracking its brand and a few competitors, €300 to €1,500 for a multi-client platform aimed at agencies. Some SEO suites include the function for €20 to €80 a month extra.
What criteria should you use to choose?
The assistants covered, support for your language, the number of questions tracked, the reading frequency and the number of repetitions per question, detection of source pages rather than brands alone, competitor tracking, and the ability to export into an existing dashboard.
Is a tool essential?
Not for a micro-business or a small company with lower stakes: a manual reading of ten to twenty questions put three times to three assistants, once a quarter, takes thirty minutes and supplies the essentials. A tool becomes useful beyond fifty or so questions, or for multi-client tracking.