How to Measure AI Search Visibility

AI search visibility is how often AI tools like ChatGPT, Perplexity and Google AI Overviews mention or cite your brand when buyers ask about your category. You measure it with four metrics: citation frequency, AI share of voice, AI-referred traffic, and your competitive citation gap. Track them across a fixed set of buyer prompts, re-run monthly, and split by platform.

What we'll cover

1. Why keyword rankings no longer tell the full story
2. What AI search visibility actually means
3. The four metrics that define AI search visibility
4. How to build a prompt set you can track
5. How to run the measurement, step by step
6. Which tools track AI search visibility
7. How to read your benchmarks
8. What to put in your monthly AI visibility report
9. A worked example: reading one prompt
10. Turn the gaps into pipeline
11. Five mistakes teams make when measuring AI visibility

Why keyword rankings no longer tell the full story

A ranking tells you where a blue link sits. It says nothing about whether you show up inside the answer a buyer actually reads.

That gap is widening fast. In the first four months of 2026, 68% of Google searches ended without a click, and for every 1,000 US searches only 276 now reach the open web. Buyers get their shortlist inside the results, or inside an LLM, before they ever land on your site.

Here is the part that ties the two worlds together. Traditional SEO is now the entry ticket to AI visibility: roughly 97% of AI Overview citations come from pages already ranking in the top 20 organic results. You cannot measure your way out of weak fundamentals. What you can do is see, clearly, whether the work is translating into presence in the answers. For the wider shift in buyer behaviour, our AI search statistics roundup has the numbers.

What AI search visibility actually means

The unit that matters for B2B SaaS is the entity, meaning your brand as a recognised thing, not a single URL. You want the model to recommend the company, then cite a page to back it up.

That distinction shapes everything downstream. It changes which prompts you track, which numbers you report, and what "winning" looks like. Fix it before you open a single tool, or you will measure page-level noise instead of brand-level presence.

The four metrics that define AI search visibility

Most teams over-complicate this. Four metrics answer the only questions that matter: does AI cite me, do I win my category, does it send traffic, and where am I losing to rivals.

If this looks like more than four numbers underneath, it is. Mike King of iPullRank breaks measurement into input, channel and performance metrics, from entity salience and bot activity through to how citations turn into sales.

Start with the four above. They are enough to run a real programme, and you can layer King's deeper input metrics on once the basics are moving.

How to build a prompt set you can track

Measurement starts with a fixed prompt set. Your numbers are only comparable month to month if the questions stay the same, so treat the set as an asset you version, not something you improvise each time.

Here is the sequence we use with clients.

1. Write 20 to 50 prompts that mirror how buyers really ask. Pull the exact wording from sales calls, your CRM, and the "people also ask" boxes in your category.
2. Balance three prompt types: informational ("what is the best tool for X"), comparison ("tool A versus tool B"), and recommendation ("recommend a platform for a mid-market RevOps team").
3. Test across buyer personas, not one generic buyer. The consultant Aleyda Solis frames AI search measurement around exactly this: quantify how much traffic LLMs send, test how your brand shows up across personas, and track where your citations come from.
4. Segment by funnel stage, so you can see whether you appear for early "what is" questions as well as late "best tool for" ones.
5. Freeze the set. Store it in a sheet, give each prompt an ID, and change it only in deliberate, logged versions.

Then weight your effort by where your buyers are. ChatGPT holds roughly 60% of the AI search market, and 47% of B2B buyers name it their preferred research LLM. That makes it the first platform to instrument, with Perplexity close behind for referral traffic. Our guide to buyer queries for AI prompt tracking goes deeper on sourcing the prompts.

How to run the measurement, step by step

You can measure AI visibility in three passes, from a free manual method to a fully instrumented one. Start with the manual pass even if you plan to buy a tool. It teaches you what good looks like.

The manual pass

Run your prompt set by hand on each platform and log the result in a sheet: prompt ID, platform, mentioned yes or no, cited yes or no, your rough position in the answer, and which rival brands were named. One person can cover a 30-prompt set across three platforms in an afternoon. Do it monthly and you already have citation frequency, a first read on share of voice, and a competitive gap list.

The GA4 pass

To capture AI-referred traffic, add a custom channel in GA4 where Medium equals referral and Source matches an AI regex (chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com), then move that rule above the standard Referral channel. Full setup and a current regex list are worth following closely, and you should refresh the pattern each quarter as new engines appear.

The tools pass

When the manual pass gets heavy, move to a dedicated tracker that runs your prompt set automatically and reports mentions, citations and share of voice by platform. Which tool to pick is next.

Which tools track AI search visibility

Two camps exist: dedicated AI-visibility platforms, and AI modules bolted onto the SEO suites you already pay for. Here is an honest comparison of the main options, with a wider roundup if you want the full field.

Also worth testing: Evertune, AthenaHQ and Otterly, all dedicated platforms that skew B2B. Whichever you trial, judge it on four things: a per-platform breakdown, entity-level share of voice, a competitive gap view, and source-level citation data, rather than one blended score.

The SEO suites are the cheapest way to start if you already use them. The dedicated platforms go deeper on prompt-level detail. Across our own campaigns we pair one broad tracker for the trend with manual spot-checks on the prompts that matter most, because no tool sees every answer.

How to read your benchmarks

Raw numbers mean little without context. Use these category benchmarks to read your AI share of voice.

Do not chase 100%. Even category leaders rarely clear 60%, because AI systems deliberately diversify the sources they cite. And always read the number per platform, since a healthy 35% on ChatGPT can hide a 5% on Perplexity that is quietly costing you referral traffic.

What to put in your monthly AI visibility report

A useful report fits on one page and shows movement, not just a snapshot. Make the trend and the next action obvious to someone who was not in the weeds with you.

Keep four things on the page.

1. The four metrics, split by platform, each shown as this month against last so the direction is visible at a glance.
2. Your top three competitive citation gaps, ranked, each with the exact prompt and the rivals winning it.
3. The actions you took last month and what moved, so the report reads as a loop, not a scoreboard.
4. One headline number for the executive view: entity-level share of voice in your priority category.

A worked example: reading one prompt

Say you sell a mid-market analytics platform and you track the prompt "best product analytics tools for B2B SaaS" across three engines. In one month you log:

• ChatGPT: two rivals named, you are mentioned but not cited.
• Perplexity: you are cited, alongside three competitors.
• Gemini: neither mentioned nor cited, two rivals named.

For this prompt your citation frequency is one of three, your share of voice is roughly one of eight brand mentions (about 12%), and your competitive gap is clear. You are absent on Gemini and uncited on ChatGPT. The action writes itself: earn a place in the comparison content those two engines lean on for this query, then re-measure next month.

Turn the gaps into pipeline

A dashboard is not the goal. More citations in the answers your buyers read is the goal.

Take your competitive gap list and map each row to a specific move. Earn a place in the buyer guides and comparison articles that LLMs lean on. Tighten the entity signals on your key pages with clear, structured, consistent descriptions of who you are. Answer the exact questions where rivals currently win. This placement-led approach is the core of MADX's AI search service, and our guide to ranking in AI Overviews covers the on-page side.

One pattern shows up again and again: the biggest jumps come from a handful of well-placed comparison mentions, not from publishing more blog posts. Re-measure monthly, watch the gap close, and let citation rate, not vanity rankings, tell you it is working.

Five mistakes teams make when measuring AI visibility

Most measurement problems trace back to the same handful of errors. Avoid these and your numbers will actually mean something.

1. Changing the prompt set every month. If the questions move, the trend is noise. Freeze the set and version it deliberately.
2. Tracking one platform. ChatGPT leads, but a blind spot on Perplexity or Gemini hides real gaps. Measure each engine on its own.
3. Counting mentions but not citations. A mention with no cited source rarely drives a click. Separate the two and report both.
4. Trusting GA4 at face value. Stripped referrers push AI sessions into direct, so an unexplained direct rise may be your AI traffic in disguise.
5. Skipping the competitive gap. Your own score without rivals in view is half a picture. The gap is where the roadmap lives.

Get the measurement honest first. Once you can see where AI leaves you out, closing the gap becomes an execution problem, not a guessing game.

Frequently Asked Questions

What is a good AI share of voice for B2B SaaS?

In most categories, 25 to 40% is competitive and anything above 40% is strong. Under 15% suggests AI rarely mentions you relative to rivals, which points to a citation gap worth closing. Very few brands exceed 60%, so treat that as a practical ceiling rather than a target.

How is AI search visibility different from SEO rankings?

SEO rankings measure the position of a blue link on a results page. AI search visibility measures whether AI answers name and cite your brand at all. You can rank first and still be left out of the synthesised answer, which is why the two need separate tracking. That said, the two are linked: most AI Overview citations come from pages already ranking in the top 20.

How often should I measure AI search visibility?

Monthly is enough for most B2B SaaS teams. AI answers shift as models update and content changes, so a stable prompt set re-run each month gives you a clean trend without chasing daily noise.

Which AI platform should B2B SaaS track first?

Start with ChatGPT, which holds the largest share of AI search and is the LLM most B2B buyers name for vendor research. Add Perplexity next, since it drives more referral traffic than its mention rate implies, then Gemini and Google AI Overviews.

Can I measure AI search visibility for free?

Yes, to a point. You can run a small prompt set manually across each platform and log the results in a spreadsheet, and you can track AI referrals in GA4 with a custom channel. Dedicated tools become worth it once you need per-platform trends and reporting across many prompts. Some, such as Ahrefs Brand Radar, offer a free tier.

Why is my AI-referred traffic showing as direct in GA4?

AI apps and in-answer browsers often strip the referrer before the click reaches your site, so GA4 has nothing to attribute and files the session under direct. Watch for unexplained rises in direct traffic to AI-relevant landing pages, and lean on citation tracking to fill the gap.