"How are we doing in AI search?" is now a standard question in monthly marketing meetings, and most teams answer it with a screenshot of a ChatGPT response. That is not reporting. It is an anecdote.
A useful AI search report needs a small, stable set of KPIs that are defined the same way every month, come from sources you can explain and connect to business outcomes. This post defines that set: visibility, citations, share of voice, sentiment and accuracy, AI referrals and assisted conversions. For each one we explain the definition, where the data comes from, how to set a target and how often to report it.
We also cover what has changed in 2026, because new first-party reports from Google, Microsoft and GA4 mean some of these KPIs no longer have to rely on third-party estimates.
Key Takeaways
- Report a short, fixed set of seven KPIs across three tiers: visibility, perception and outcomes.
- Use first-party data where it exists: Search Console's Generative AI report, Bing's AI Performance report and GA4's AI Assistant channel.
- Set targets relative to your own baseline (for example, the first three months), not to industry averages that rarely apply.
- Report visibility and referrals monthly; sentiment, accuracy and competitive share quarterly.
- Always show AI referrals next to conversions, and state the known gaps in attribution.
The Three Tiers of AI Search KPIs
Think of AI search measurement as a funnel. Each tier answers a different question:
- Visibility: Do AI platforms show or mention us?
- Perception: How do they describe us, and is it accurate?
- Outcomes: Does it bring visitors, leads and revenue?
Most reports overweight tier one because it is easiest to measure. A balanced report gives tier three at least equal space.
The KPI Set
| KPI | Definition | Main source | Cadence |
|---|---|---|---|
| 1. Google AI impressions | Times links to your site were shown in AI Overviews and AI Mode | Search Console Generative AI report | Monthly |
| 2. Copilot / Bing AI citations | Times your pages were cited in Copilot and Bing AI summaries | Bing Webmaster Tools AI Performance | Monthly |
| 3. Prompt mention rate | % of tracked prompts (averaged over several runs) where your brand is mentioned | Prompt tracker or manual sample | Monthly |
| 4. AI share of voice | Your mentions or citations as a share of all tracked brands' mentions for the same prompts | Prompt tracker | Quarterly (monthly for competitive markets) |
| 5. Sentiment and accuracy | % of answers that describe you positively and without factual errors | Manual review of a sample | Quarterly |
| 6. AI referral sessions and engagement | Sessions, engaged sessions and engagement rate from the AI Assistant channel | GA4 | Monthly |
| 7. AI-influenced conversions | Key events from AI referrals plus self-reported "found via AI" leads | GA4, CRM, form question | Monthly |
“AI Assistant is the channel by which users arrive at your site from sources like ChatGPT, Gemini, Deepseek, Copilot, or Grok.”
— Google Analytics Help, default channel group definitions, [GA4] Default channel group
One important detail from the same documentation: the AI Assistant channel excludes Google's AI Overviews and AI Mode, which are counted in Organic Search. So KPI 6 covers assistants, while KPI 1 covers Google's AI features. Do not add them together or describe either as "all AI traffic".
How to Set Targets Without Inventing Benchmarks
There are no reliable universal benchmarks for AI visibility. Platforms differ, prompt sets differ and answers vary between runs. Instead:
- Baseline for three months. Collect every KPI without targets. Note the range, not just the average.
- Set direction targets first. For example, "increase prompt mention rate in the 'pricing' cluster" or "reduce factual errors in brand answers to zero".
- Add numeric targets from your own baseline. Aim for an improvement that is bigger than your normal month-to-month variation, otherwise you are just measuring noise.
- Tie outcome targets to the business plan. If AI referrals are a small share of traffic, set the outcome target on lead quality and conversion rate rather than volume.
- Review targets quarterly. Platforms change fast; a feature launch can move numbers overnight.
Monthly vs Quarterly Reporting
| Report | Include | Audience | Format |
|---|---|---|---|
| Monthly snapshot | KPIs 1, 2, 3, 6, 7 with trend vs previous month and baseline; top cited pages; notable changes | Marketing team | One-page dashboard plus 3–5 bullet commentary |
| Quarterly review | All seven KPIs; share of voice by competitor; sentiment and accuracy review; actions taken and results | Leadership | Short deck or written memo |
| Ad hoc alert | Sudden drop in citations, factual error found, platform change | Owner of the issue | Message with screenshot and next step |
Semrush's September 2026 reporting template takes a similar approach, suggesting most metrics monthly while sentiment, backlinks and competitive share of voice suit quarterly reporting. We show how to build the monthly view in our guide to an AI search dashboard in Data Studio.
Measuring the Hard-to-Measure KPIs
Sentiment and accuracy
Take 20–30 brand prompts ("Is [brand] good for…", "[brand] vs [competitor]", "[brand] pricing") each quarter. Record each answer and score it: positive, neutral or negative, and accurate or inaccurate. Log every factual error with the source the AI cited, then fix it at the source. Our AI sentiment analysis guide covers scoring methods.
AI-influenced conversions
Analytics will undercount AI influence, because many people read an AI answer and later search your brand or type your URL. Two practical fixes:
- Add "How did you hear about us?" to forms with an AI option such as "ChatGPT or another AI assistant". See self-reported attribution for AI.
- Watch branded search trends in Search Console alongside AI visibility, as explained in branded search as an AI proxy and our guide to branded vs non-branded keywords.
Traffic quality
Report engagement rate and conversion rate for AI referrals next to organic search. If AI visitors behave differently, that is worth knowing; our post on AI traffic quality explains how to compare fairly.
Common Mistakes
- Changing definitions mid-year. If you change the prompt set or the share-of-voice formula, restart the baseline.
- Reporting a single run as a fact. Prompt-based KPIs need several runs per period.
- Mixing Google AI impressions with clicks. The Generative AI report has impressions only; clicks are inside the standard Performance report.
- Borrowing benchmarks from vendor studies. Use them for context, labelled and dated, never as your targets.
- Too many KPIs. Twenty metrics means nobody reads the report. Seven is plenty.
- No commentary. Every report should say what changed, why you think it changed and what you will do next.
For how these KPIs fit into wider SEO measurement, see our SEO KPIs guide and our post on measuring the ROI of AI search visibility.
Related Guides
- Detecting AI Agent Visits in Your Analytics
- AI Search for Local Businesses: A Practical Starter Plan
- AI Search for Restaurants: Getting Recommended for "Where Should We Eat?"
Frequently Asked Questions
What are the most important AI search KPIs?
A balanced set covers visibility (Google AI impressions, Bing citations, prompt mention rate), perception (share of voice, sentiment and accuracy) and outcomes (AI referral sessions and AI-influenced conversions).
Is there a benchmark for a good AI share of voice?
Not a reliable universal one. Share of voice depends on your market, prompt set and competitors. Set targets from your own three-month baseline and track the trend.
Does GA4 track AI Overviews traffic?
GA4 includes AI Overviews and AI Mode clicks in Organic Search, so you cannot separate them there. The AI Assistant channel covers assistants such as ChatGPT, Gemini, Copilot and Claude.
How often should AI search KPIs be reported?
Monthly for visibility and referrals, quarterly for share of voice, sentiment and accuracy, with ad hoc alerts for big drops or factual errors.
How do I show AI search ROI to leadership?
Pair visibility trends with outcomes: AI referral conversions, self-reported AI leads and branded search growth, and be clear about what analytics cannot capture.
Report Less, Explain More
Seven well-defined KPIs, reported consistently and explained honestly, will tell you far more than a wall of AI screenshots. Start your baseline this month, set targets next quarter and keep outcomes front and centre. If you would like a ready-made AI search reporting setup for your team or clients, see our client marketing reports guide or get a quote from us.
References
- [GA4] Default channel group – Google Analytics Help
- What's new in Google Analytics – Google Analytics Help
- Generative AI performance report (Search) – Search Console Help
- Introducing AI Performance in Bing Webmaster Tools Public Preview – Bing Webmaster Blog (February 2026)
- How to measure AI search visibility: KPIs & reporting – Semrush Blog (May 2026)
- Create an AI Brand Visibility Report – Semrush Blog (September 2026)
