Should I track ChatGPT with or without browsing?

Track both. Browsing-enabled answers reflect fresh retrieval; non-browsing answers reflect training-data recall. They tell different stories, and your brand can be visible in one and absent from the other.

Can Google Search Console show my ChatGPT visibility?

No. The generative-AI performance reports Google announced in June 2026 cover AI Overviews and AI Mode, and are rolling out to a subset of sites. They are a valuable free baseline for Google's AI surfaces, but ChatGPT visibility still requires prompt sampling or a dedicated tracking tool.

Do I need a paid tool to get started?

No. A spreadsheet handles up to roughly 30 prompts, and free entry points exist — AthenaHQ has a free Essential tier, and HubSpot's free AEO Grader runs a one-time analysis across ChatGPT, Perplexity, and Gemini. Paid tools earn their keep once you need recurring, multi-engine tracking at scale.

What counts as a good share of voice?

There is no universal benchmark — SoV depends on how many competitors share your category and how you build the prompt set. What matters is the trend on a consistent prompt set and how you compare against named competitors, segment by segment.

How to Track ChatGPT Visibility: A Step-by-Step Guide

Q: How many samples per prompt do I need?

Three to five samples per prompt per run is a reasonable balance between signal and cost. Ahrefs has documented that identical prompts can produce different responses, so a single sample can misstate your visibility; averaging repeated samples stabilizes the trend.

Why ChatGPT visibility needs its own tracker

You cannot read ChatGPT visibility off a Google rank tracker. An Ahrefs test of 3,311 head terms found ChatGPT's citations overlapped Google's top ten results only 31.8% of the time at the domain level and just 10% at the exact-URL level — far below Perplexity, where overlap was 80.58% and 65.07%. A brand that dominates page one of Google can be invisible in ChatGPT answers, and vice versa. Since ChatGPT shapes shortlists before a buyer ever visits a website, the only way to know where you stand is to measure ChatGPT's answers directly: define a prompt set, sample it on a schedule, tag mentions, and compute share of voice. The five steps below walk through exactly that.

Step 1 - Define your prompt set

Build 30-100 prompts that reflect how buyers actually ask ChatGPT about your category. Mix category prompts ('best X for Y'), comparison prompts ('X vs Y'), alternative prompts ('alternatives to X'), and use-case prompts ('how do I do Z'). Weight the set toward opening questions rather than mid-thread follow-ups: Profound's analysis of roughly 730,000 cited US-English ChatGPT conversations from October through December 2025 found citation incidence was highest at the first turn — 12.6%, falling to 4.5% by turn ten and 3% by turn twenty. The first exchange is where sources get surfaced, so phrase prompts the way a buyer would start a conversation.

Pull language from sales calls and support tickets
Include competitor names
Cover the full funnel: awareness -> consideration -> decision
Phrase prompts as conversation openers, where citations concentrate

Step 2 - Sample consistently, and more than once

Run each prompt on a consistent cadence — weekly for active tracking, monthly for a baseline — in fresh sessions so personalization and memory do not skew results. Critically, never rely on a single sample per prompt: Ahrefs has documented that identical prompts can produce different responses across repeated runs. One response is an anecdote, not a measurement. Collecting three to five samples per prompt per run smooths most of that variance, and averaging across runs turns noisy individual answers into a stable trend line. If your visibility number jumps around week to week, increase samples per prompt before concluding anything changed.

Step 3 - Tag mentions

For each response, tag whether your brand is mentioned, the position of the mention, sentiment, and which competitors also appear. Normalize brand aliases so 'Acme', 'Acme Inc', and 'acme.com' collapse to one entity. Log cited URLs too — knowing which sources ChatGPT leans on for each prompt is what makes the data actionable later.

Step 4 - Calculate share of voice

Share of voice (SoV) is your brand's mentions divided by total brand mentions — yours plus competitors' — across all sampled responses in a run. Track it over time and segment by prompt type to see where you are strong versus weak.

SoV = your brand mentions / (your mentions + competitor mentions)
      across all sampled responses in the run

Segment by prompt type (category, comparison, alternatives, use-case)
to find where you win and where you are absent.

A worked example (illustrative)

Here is a filled-in example — the numbers are illustrative, not measured data. Suppose you track 20 prompts across four types (five each) and collect three samples per prompt, for 60 responses per run. Reading the results like a table, one row per prompt type:

Category prompts, 15 responses: your brand 3 mentions, competitors 21 - 24 total, 12.5% SoV
Comparison prompts, 15 responses: your brand 9 mentions, competitors 15 - 24 total, 37.5% SoV
Alternatives prompts, 15 responses: your brand 6 mentions, competitors 18 - 24 total, 25.0% SoV
Use-case prompts, 15 responses: your brand 2 mentions, competitors 10 - 12 total, 16.7% SoV
Overall: 20 of 84 brand mentions - 23.8% share of voice

Step 5 - Act on the data

In the illustrative example above, the brand wins comparison prompts (37.5%) but nearly vanishes on category prompts (12.5%) — and category prompts are where buyers who have never heard of you start. Closing that kind of gap usually means earning coverage in the sources ChatGPT cites for those prompts, not just publishing more yourself: Profound's citation-category analysis found owned sources supplied only 4.3% of citations for category prompts. And the coverage that works is editorial, not paid — Muck Rack's May 2026 analysis of more than 25 million links across ChatGPT, Claude, and Gemini found earned media made up 84% of links while paid or advertorial content contributed 0.3%. Sentiment issues, by contrast, resolve through messaging fixes and review response.

Free first-party data: Google Search Console's generative-AI reports

In June 2026 Google announced generative-AI performance reports in Search Console, covering AI Overviews and AI Mode with breakdowns by page, country, device, and date. It is rolling out to a subset of sites, so check whether yours has access. This is a meaningful free, first-party measurement surface for AI search — worth adopting immediately, since Google's own AI-search guidance notes that third-party tools do not have access to Google's internal search data. Two caveats keep it in perspective: availability is still limited, and it reports Google's AI surfaces, not ChatGPT. Treat it as the free baseline for Google AI visibility that runs alongside — not instead of — the ChatGPT prompt-sampling program described above.

Tools that automate this

For manual programs, a shared spreadsheet works up to roughly 30 prompts. Beyond that, dedicated platforms automate sampling, tagging, and reporting. List prices below are from vendor pricing pages checked in July 2026.

Profound - Starter $99/month billed yearly (ChatGPT only, 50 prompts); Growth $399/month billed yearly adds Perplexity and Google AI Overviews
Otterly AI - Lite $29/month for 15 prompts; base engines are ChatGPT, Google AI Overviews, Perplexity, and Copilot; 7-day trial, no permanent free plan
Peec AI - Starter $95/month (50 prompts, one project); self-serve plans choose three of six engines
AthenaHQ - free Essential tier with 300 credits; Starter $295/month
HubSpot AEO - $50/month, or $45/month paid annually, for 25 daily-tracked prompts across ChatGPT, Perplexity, and Gemini (beta)
Semrush AI Visibility - standalone Base plan $99/month per domain billed annually, with 25 custom prompts