How to Measure AI Search Visibility for a SaaS Company
Amit Malvi
Founder, SearchAxe. Former Senior SEO Manager, Invideo AI
Published Updated 11 min read
Quick answer
If you want to know how to measure AI search visibility for a SaaS company, start with rates, not rankings. Build a fixed set of prompts from real buyer questions, run each one several times on every engine, and record the share of runs that mention, recommend and cite you. Add share of voice against named competitors and a log of misdescriptions, then connect it to AI referral traffic, self-reported attribution and pipeline in your CRM.
On this page
- How to measure AI search visibility for a SaaS company, in five layers
- Which prompts should a SaaS prompt set include?
- What should you record for each prompt and engine?
- How do you calculate share of voice in AI answers?
- How do you measure accuracy and sentiment in AI answers?
- How do you connect AI visibility to referral traffic and pipeline?
- Should you track AI visibility manually or with a tool?
- What should an AI visibility scorecard look like?
Key takeaways
- Measure rates across repeated runs, not positions, because AI engines rarely return the same brand list twice for the same prompt.
- Build the prompt set from sales calls and lost-deal notes, grouped by intent, and keep it fixed so month-on-month changes mean something.
- Track four rates per prompt and engine: mentioned, recommended, cited and described correctly, plus share of voice against a fixed competitor set.
- AI referrals are a small slice of traffic, 0.32% of web traffic in SE Ranking's June 2026 study, so add a self-reported attribution field to catch buyers who never click.
- Free native reports exist only for Google and Microsoft surfaces, so most SaaS teams combine a manual or tool-run prompt sheet with Search Console and Bing Webmaster Tools.
You measure AI search visibility with a fixed prompt set, repeated runs on each engine and a short scorecard that ends in pipeline, not screenshots. If you run marketing or growth at a seed to Series B SaaS or AI company and the board has started asking whether ChatGPT recommends you, this is the framework we'd hand you.
How to measure AI search visibility for a SaaS company, in five layers
Measure it in five layers, each built on the one before. First a prompt set that mirrors how your buyers ask. Then per-engine rates for mentions, recommendations and citations. Then share of voice against named competitors. Then accuracy and sentiment. Then the commercial layer: AI referral sessions, self-reported attribution and pipeline.
The reason for rates rather than ranks is variance. In a SaaS category with many credible tools, the shortlist an engine gives changes from run to run. In SparkToro and Gumshoe's repeated-run research, reported by Search Engine Journal in January 2026, how often a brand appeared across many runs was more consistent than where it appeared in any one answer. So the unit of measurement is the share of runs, and a position number from a single run should be ignored. If you need the underlying definition first, read what AI visibility means and how it differs from search rankings.
Which prompts should a SaaS prompt set include?
Include the questions buyers ask before they shortlist, written the way they type them into an AI engine, grouped by intent. Pull them from sales call recordings, demo request forms and lost-deal notes, not from your keyword tool, because prompts are longer and carry the buyer's context: team size, stack, budget and the job they are hiring the tool for.
| Intent group | What it captures | Example prompt |
|---|---|---|
| Category shortlist | Who gets named when the buyer has not picked a vendor | "What's the best AI note taker for a 20-person sales team" |
| Use case | Whether you are tied to the job, not just the category | "How can I get call notes into HubSpot automatically" |
| Alternatives | Whether you appear when a buyer is leaving a rival | "Alternatives to Gong for a seed-stage startup" |
| Head-to-head | How you are compared when you are already on the list | "Your product vs Fireflies for sales calls" |
| Fit and pricing | Whether the answer gets your plans and audience right | "Is your product affordable for a team under 10" |
| Brand | What the engine says when asked about you directly | "What is your product and who is it for" |
Build the prompt set in an afternoon
- Pull 20 recent sales calls and demo formsCopy the exact phrases buyers used to describe their problem, because those words are what they will type into an engine.
- Write prompts for each intent groupWeight the set toward shortlist, use case and alternatives prompts, since those decide whether you are considered at all.
- Fix the competitor listName the four to six rivals that appear in your deals, so share of voice is always calculated against the same field.
- Freeze the set for at least a quarterChanging prompts every month makes trends meaningless; add new prompts as a separate group instead of swapping old ones out.
Find out which AI engines recommend you
A free consultation on where you appear today across AI engines and search.
Schedule a Free ConsultationWhat should you record for each prompt and engine?
Record four yes-or-no fields per run and turn them into rates: whether the brand was mentioned, whether it was recommended for the stated need, whether one of your URLs was cited (and which one), and whether the description was correct. Run each prompt several times on each engine, in a clean session with no saved memory, and divide the yeses by the runs.
- Mention rate: runs that name you, divided by total runs for that prompt and engine.
- Recommendation rate: runs that suggest you for the buyer's need, which is the number closest to revenue.
- Citation rate and cited URLs: which of your pages, and which third-party pages about you, the engine links.
- Accuracy rate: runs where category, audience, pricing model and core features are all described correctly.
- Competitor mentions: the same fields for each rival on your fixed list, which feed share of voice.
Cover every engine your buyers use: ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews, Google AI Mode and Copilot. They do not move together. Ahrefs found in its August 2025 study of 15,000 prompts that around 8% of URLs cited by ChatGPT, Gemini and Copilot ranked in Google's top 10 for the same prompt. For Perplexity the figure was 28.6%. A single-engine number hides most of the picture.
How do you calculate share of voice in AI answers?
Share of voice is your mentions divided by the mentions of every brand on your fixed list, across the same prompts, engines and runs. If your product is named in 12 runs and the five tracked rivals are named 48 times between them, your share of voice is 12 out of 60, or 20%. Calculate it per intent group, because a strong brand-prompt number can hide a weak shortlist number.
Microsoft now publishes a citation version of the same idea for its own surfaces. Its June 2026 Bing Webmaster Tools update defines citation share as the percentage of citations attributed to your site out of all citations shown for the same grounding query, across Copilot, Bing and select partner experiences. For the formula, worked examples and benchmarking, see what share of voice in AI search is and how to calculate it.
How do you measure accuracy and sentiment in AI answers?
Write a one-page fact sheet of what must be true in any description of your product, then score every answer that mentions you against it. For SaaS the fields that go wrong most are category, target customer, pricing model, integrations and whether a feature still exists. Log each error with the engine, the exact wording and the sources cited in that answer.
The cited sources are the point. A misdescription usually traces back to a page the engine retrieved: an outdated roundup, an old pricing page, a review from before a pivot. Fixing the source, or publishing a clearer page of your own, is what changes the answer. Sentiment works the same way. Score it as positive, neutral or negative per answer, and when it turns negative, read the sources rather than arguing with the output.
How do you connect AI visibility to referral traffic and pipeline?
Track three things in your analytics and CRM: AI referral sessions, signups and demos from those sessions, and self-reported attribution for buyers who never clicked. The first two show the click path. The third catches buyers who read an answer, then typed your name into Google or Bing a day later.
OpenAI's publishers FAQ says ChatGPT adds utm_source=chatgpt.com to referral links, so those sessions are easy to isolate in GA4. Clicks from Google AI Overviews and Google AI Mode are different. Google says they are counted inside your overall Search Console traffic under the Web search type, and in GA4 they arrive as ordinary Google organic visits, so you cannot split them out there. Our GA4 guide to AI referral traffic has the channel group and regex for the rest.
Small volume does not mean small value, but one company's data is not a benchmark. Ahrefs reported in June 2025 that AI search sent 0.5% of its traffic and 12.1% of its signups over 30 days, and said plainly that this was its own experience. Benchmark your own conversion rate by source. Add a required "How did you hear about us" field to signup and demo forms with "ChatGPT or another AI assistant" as an option, and pass it into the CRM so pipeline can be reported by source. Our pipeline measurement guide covers the attribution model.
Should you track AI visibility manually or with a tool?
Start manually, then buy a tool when the prompt set outgrows an afternoon a month. A manual sheet forces you to read the answers, which is where misdescriptions and competitor sources show up. Tools add volume, daily sampling and history, but they sample answers the same way you would and cannot see inside the engines.
| Method | What it covers | Strength | Limit |
|---|---|---|---|
| Manual prompt sheet | Any engine you can open in a browser | You read every answer and its sources | Few runs per prompt, so small samples |
| Prompt tracking tools, such as Ahrefs Brand Radar or Semrush's AI Visibility Toolkit | The engines each vendor supports, on custom prompts | Daily runs, history, competitor share of voice | Sampled answers, and coverage differs by vendor |
| Search Console generative AI report | Google AI Overviews and Google AI Mode | First-party impressions by page, country and device | Impressions only, no clicks or unlinked mentions |
| Bing Webmaster Tools AI Performance | Copilot, Bing AI summaries and select partners | First-party citation counts and grounding queries | Citations only, no clicks, Microsoft surfaces only |
Google's generative AI optimization guide adds a useful warning: no third-party tool has access to its internal ranking or AI systems. Treat tool numbers as samples. Ahrefs Brand Radar runs custom prompts across several engines, and Semrush's AI Visibility Toolkit reports share of voice and sentiment. We compare the options in AI visibility tracking tools compared for startups, and the no-cost starting point is in how to check whether ChatGPT and Perplexity recommend your product without paid tools.
What should an AI visibility scorecard look like?
One page, nine rows, reported monthly against the baseline you recorded before any work started. Visibility rows come first because they move first; pipeline rows come last because they lag. Every number has a defined source so the next person can reproduce it.
| Metric | How it is calculated | Data source | Cadence |
|---|---|---|---|
| Mention rate | Runs naming you divided by total runs, per engine and intent group | Prompt sheet or tracking tool | Monthly |
| Recommendation rate | Runs recommending you for the stated need divided by total runs | Prompt sheet or tracking tool | Monthly |
| Share of voice | Your mentions divided by mentions of all tracked brands | Prompt sheet or tracking tool | Monthly |
| Cited pages | Unique URLs of yours cited in answers | Prompt sheet, Search Console, Bing Webmaster Tools | Monthly |
| Accuracy rate | Answers matching your fact sheet divided by answers mentioning you | Prompt sheet | Monthly |
| Google AI impressions | Impressions in Google AI Overviews and Google AI Mode | Search Console generative AI report | Monthly |
| AI referral sessions | Sessions from AI engine referrers and utm_source=chatgpt.com | GA4 | Monthly |
| AI-attributed signups and demos | Conversions from AI referrals plus self-reported AI answers | GA4 and form field | Monthly |
| AI-influenced pipeline | Opportunity value where source or self-report is an AI engine | CRM | Quarterly |
Report the scorecard monthly and judge it on three-month trends; the sampling cadence and logging setup are in how to track brand mentions across ChatGPT, Perplexity, Gemini and Google AI. This is the reporting an AI visibility agency should be producing for you, whoever does the work.
Frequently asked questions
What is the most important AI visibility metric for a SaaS company?
Recommendation rate on shortlist and alternatives prompts. It is the share of repeated runs in which an engine suggests your product for the buyer's stated need, and it sits closest to revenue. Mention rate shows category association and citation rate shows whether your pages are usable, but a buyer who asks for the best tool for their team acts on the recommendation. Track it per engine and per intent group, against a fixed competitor list.
How should you report AI visibility to a board?
Show three numbers against the baseline you recorded before any work started: share of voice on shortlist prompts, recommendation rate for the buyer's stated need, and AI-attributed pipeline from referral sessions plus the self-reported field in your CRM. Report the first two monthly and pipeline quarterly, and judge all three on three-month trends, since single months swing. Keep the full scorecard as an appendix so anyone can trace each number to its source.
Can you see AI search visibility in Google Analytics?
Only the clicks. GA4 shows sessions that arrive from AI engines, and ChatGPT tags its referral links with utm_source=chatgpt.com, according to OpenAI. It cannot show answers that named you without a click, and Google counts clicks from Google AI Overviews and Google AI Mode inside normal Search traffic. Pair GA4 with a prompt sheet for visibility and a self-reported attribution field for buyers who never clicked.
Is share of voice better than mention rate?
They answer different questions, so report both. Mention rate tells you how often you appear on your own prompts. Share of voice tells you how much of the conversation you own compared with a fixed set of rivals, and it can fall even when your mention rate holds, because a competitor is gaining. For a board update, share of voice on shortlist prompts is usually the clearest single number to show.
When is an AI visibility tool worth paying for?
When the manual sheet takes more than an afternoon a month, when you need daily sampling to see the effect of a launch, or when you track more than a handful of competitors. Before that, a manual sheet plus Search Console's generative AI report and Bing Webmaster Tools' AI Performance report covers the basics. Google notes that no third-party tool can see inside its ranking or AI systems, so read tool output as a sample.
