How to Measure GEO: Tracking AI Visibility for B2B SaaS

Most B2B SaaS marketing leaders can tell you their organic traffic numbers, keyword rankings, and conversion rates from search. Ask them how often ChatGPT, Perplexity, or Gemini recommends their product, and you’ll get a blank stare. That’s the problem. Buyers increasingly use AI tools to research software categories, compare vendors, and form a view before they ever engage with a sales team. If your company isn’t in those answers, you’re not on the shortlist. Measuring your AI visibility, often called Generative Engine Optimisation (GEO), requires a different set of tools and thinking than traditional SEO. The metrics aren’t as mature, the data isn’t as clean, and some of it still requires manual work. But the signals are there if you know where to look. This piece walks through what you can track right now, what’s still out of reach, and how to build a practical measurement programme that gives your team real numbers to work with.

Why measuring GEO is different from SEO

SEO measurement is built on a well-established infrastructure. Google Search Console gives you impressions and clicks. GA4 tracks sessions and conversions. Rank trackers show you positions for target keywords. You’ve had two decades to refine this.

GEO has none of that infrastructure. AI models don’t send referral data the same way a search engine does. When ChatGPT mentions your product in a response, there’s no impression logged, no click-through rate, and no ranking position in the traditional sense. The “result” is a paragraph of synthesised text, not a list of ten blue links.

The unit of measurement changes too. In SEO, you optimise for keywords. In GEO, you’re optimising for entity recognition: whether AI models understand what your company does, which category you belong to, and whether they consider you credible enough to cite. GEO metrics tend to track brand mentions, citation frequency, and sentiment accuracy rather than clicks and impressions.

There’s also the problem of opacity. Google publishes documentation about how its algorithm works. OpenAI, Anthropic, and Google’s Gemini team don’t publish equivalent guidance for how their models select which brands to mention. You’re measuring outputs without full visibility into the process that generates them.

This doesn’t mean measurement is impossible. It means you need to accept a different level of precision and build your tracking programme around what’s observable today.

Track LLM referral traffic in GA4

The most concrete GEO metric you can capture right now is referral traffic from AI tools. When someone clicks a link in a ChatGPT, Perplexity, or Gemini response, that visit shows up in your analytics. The challenge is identifying it correctly.

By default, GA4 often lumps AI referral traffic into “direct” or “unassigned” buckets. You need to create custom channel groupings to isolate it. Start by identifying the referral domains. Traffic from ChatGPT typically shows a referrer of chatgpt.com or chat.openai.com. Perplexity traffic comes from perplexity.ai. Gemini traffic arrives from gemini.google.com.

Here’s a practical setup:

1. In GA4, go to Admin, then Channel Groups, and create a new custom channel group called “AI Referrals.”
2. Add rules matching the source/medium to known AI referral domains: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.microsoft.com.
3. Apply this channel group to your acquisition reports and set up a weekly check.

The volume will likely be small compared to organic search. That’s fine. What matters is the trend line. If you’re running a GEO programme alongside your SEO work, you should see this traffic grow over time. Track which landing pages receive AI referral visits, because that tells you which content AI tools are citing. AI search traffic has been growing at roughly 50% quarter-on-quarter through 2025 and into 2026, so even modest numbers now could compound quickly.

One caveat: not all AI interactions generate clicks. Many users get their answer directly from the AI response without visiting any source. Your referral traffic is the visible tip of a larger iceberg.

Monitor AI brand mentions and citations

Since most AI interactions don’t produce a click, you need to track whether AI tools mention your brand at all. This is where manual monitoring and a few emerging tools come in.

The simplest approach is to run structured queries across ChatGPT, Perplexity, Gemini, and Claude on a regular schedule. Pick 15 to 20 prompts that reflect how your buyers research your category. For a B2B SaaS company selling, say, compliance software, those prompts might include “best compliance software for mid-market companies,” “alternatives to [competitor name],” and “how to automate SOC 2 reporting.”

Run each prompt monthly and record whether your brand appears, where it appears in the response (first mention, middle, or end), and whether it’s linked to a source. Store this in a simple spreadsheet or tracking tool. Over time, you’ll build a dataset showing your citation frequency and how it changes as you publish new content or earn new third-party mentions.

Several GEO-specific monitoring tools have emerged in 2025 and 2026. Platforms from specialist GEO agencies and tool providers now offer automated tracking of brand mentions across multiple AI models. These tools save time but aren’t cheap, so for most teams under £20M ARR, a manual process supplemented by a VA or junior marketer is a reasonable starting point.

At Gripped, we run a GEO audit as part of our work with B2B SaaS clients. That audit covers how ChatGPT, Perplexity, Gemini, and Claude currently describe the company and its category, what sources they pull from, and where competitors get cited instead. This gives teams a baseline to measure against.

Check how accurately AI describes your company

Being mentioned is only half the battle. The other half is whether the AI gets it right. Inaccurate descriptions can actively harm your pipeline if buyers form the wrong impression of what you do or who you serve.

Run your brand name through each major AI model and evaluate the response against your actual positioning. Look for specific errors: wrong product category, outdated pricing, incorrect feature descriptions, or confusion with a similarly named company. If you sell an enterprise product and the AI describes you as a “free tool for small teams,” that’s a problem you need to fix.

Create a simple accuracy scorecard. For each AI model, rate the response on three dimensions: category accuracy (does it place you in the right market?), feature accuracy (does it describe what your product actually does?), and audience accuracy (does it identify your target buyer correctly?). Score each on a 1-to-5 scale and track quarterly.

When you find inaccuracies, the fix isn’t to complain to OpenAI. It’s to ensure your web presence gives AI models consistent, correct information to train on. That means your homepage, product pages, G2 profile, LinkedIn company page, and any third-party review sites all describe your company in consistent terms. Teams that fix their structured data and align their messaging across sources tend to see AI descriptions improve over subsequent model updates.

This is where GEO and SEO overlap. Strong on-site content, consistent entity descriptions, and authoritative third-party citations serve both channels. GEO isn’t a replacement for SEO; strong SEO foundations are an input to GEO.

Measure your share of voice in AI answers

Share of voice (SOV) in AI answers tells you how often your brand appears relative to competitors for the queries that matter. This is arguably the most strategic GEO metric for B2B SaaS companies, because it maps directly to whether buyers will consider you.

To measure it, take your list of 15 to 20 category prompts and track not just whether you appear, but who else appears. If you run the prompt “best project management tools for remote engineering teams” and the AI mentions five products, your SOV for that query is 20%. If a competitor appears in four out of five AI tools and you appear in two, they have a structural advantage in buyer perception.

Build a competitive matrix:

• Rows: your target prompts (the questions your buyers ask)
• Columns: each AI model (ChatGPT, Perplexity, Gemini, Claude)
• Cells: which brands are mentioned, in what order

Run this monthly. The patterns will tell you where you’re strong, where you’re absent, and which competitors dominate. Some GEO experts recommend tracking SOV as your primary north-star metric for generative engine visibility, and we’d agree with that for most B2B SaaS companies.

The maths here is straightforward. If you appear in 30% of relevant AI responses this quarter and 45% next quarter, your GEO programme is working. If your competitor’s SOV drops from 60% to 40% over the same period, that’s a meaningful competitive shift. These numbers are more useful than raw referral traffic because they reflect the full scope of AI-mediated buyer research, not just the fraction that generates a click.

What you can’t measure yet

Honesty matters here. There are significant gaps in GEO measurement that no tool or process can fully close in 2026.

You can’t measure total AI impressions. Unlike Google, which tells you how many times your page appeared in search results, AI tools don’t report how many times your brand was mentioned across all user queries. Your manual prompt tracking captures a sample, not the population.

You can’t attribute pipeline to AI mentions with the same confidence you can attribute it to a paid search click with a GCLID. If a buyer asks ChatGPT about your category, reads the response, and then visits your site directly two weeks later, that journey is invisible in your CRM. There’s no UTM parameter, no referral cookie, no attribution trail.

You also can’t control the training data cycle. AI models update on their own schedules. Content you publish today might not influence AI responses for weeks or months. This makes it difficult to run the kind of rapid test-and-iterate cycles that work in paid media or even SEO.

What you can do is build a measurement stack that combines the observable signals: referral traffic, citation frequency, accuracy scores, and share of voice. Taken together, these give you a directional picture. Gripped’s approach with B2B SaaS clients is to run GEO tracking alongside SEO and paid reporting in 30-day sprint cycles, so the team always has fresh data to act on, even if that data is less precise than what they’re used to from other channels.

Don’t let the gaps stop you from measuring what’s available. The companies that start tracking now will have six to twelve months of trend data when better tools arrive.

Common questions

How do I see ChatGPT traffic in GA4?

Create a custom channel group in GA4 that captures referrals from chatgpt.com and chat.openai.com. Go to Admin, select Channel Groups, and add a new group with rules matching these source domains. You can extend this to include perplexity.ai, gemini.google.com, and claude.ai to capture traffic from all major AI tools in one view. Once set up, check your acquisition reports weekly. The numbers will be modest at first, but they give you a clean signal of which pages AI tools are linking to and how that traffic behaves on your site. Compare bounce rates and time on page against your organic search traffic to understand whether AI-referred visitors are genuinely engaged or just passing through.

What is a good GEO benchmark?

There’s no industry-standard benchmark yet, which is part of what makes GEO measurement tricky. For B2B SaaS companies in the £2M to £20M ARR range, a reasonable starting target is appearing in at least 25% of relevant AI responses for your core category queries. If you’re currently at zero, getting to consistent mentions in one or two AI models within 90 days is a solid first milestone. Share of voice above 30% for your primary category puts you in a strong position relative to most competitors, who aren’t tracking this at all. The real benchmark is your own trend line: are you appearing more often, more accurately, and in more AI tools than you were last quarter?

Building your GEO measurement programme

Tracking AI visibility for B2B SaaS isn’t about finding one perfect metric. It’s about combining referral traffic, citation monitoring, accuracy checks, and share of voice into a reporting cadence your team can sustain. Start with monthly manual checks, automate where tools allow, and focus your energy on the queries your buyers actually use.

The companies that build this measurement habit now will have a genuine advantage as AI-mediated buying becomes the norm. If your team is stretched and you need help building a GEO programme alongside your existing SEO and demand generation work, Gripped works exclusively with B2B SaaS and tech companies on exactly this kind of challenge. Get your free growth audit to see where your AI visibility stands today and what it would take to improve it.