Four numbers frame this guide. An AI repeats the same brand list less than 1 time in 100, so a rank report is noise (section 3). ChatGPT names brands 3.2 times more often than it links them, so counting links misses influence (section 3). AI referrals converted 90% higher than non-AI retail traffic in September 2026, so the channel earns a place on the report (section 5). And 55.9% of AI-influenced visits arrive through search, which is why a referral report undersells the channel and why the loop has to close at the first payment (section 4).
What does a CMO dashboard for AI search look like?
A CMO dashboard for AI search is one scorecard with three tiers: leading signals, quality guardrails and lagging outcomes. The sample below tracks 11 metrics, each with a target, current value, trend, status and the action it triggers. Good AI visibility tracking fills every row. The sections after it explain each row.
| # | Metric & scope | How often | Target | Current, month to date | Change, month over month | Status | Action triggered |
|---|---|---|---|---|---|---|---|
| Tier 1 · Leading Engine Signals (Early Visibility & Crawl Health) | |||||||
| 1 | Bot Crawl VolumeAI search crawlers (GPTBot, Perplexity) | Weekly | > 10,000 | 14,280 | +18.4% | HEALTHY | Monitor CDN logs; keep AI search bots allowed |
| 2 | Citation ShareFrozen 50-prompt buying corpus | Weekly | > 25.0% | 28.4% | +4.2% | ON TRACK | Expand category comparison & integration docs |
| 3 | Share of Voice vs Named RivalsPrimary 3 SaaS competitors | Weekly | > 1.50x | 1.82x | +0.32x | OUTPERFORMING | Defend tier-1 feature prompts against competitor shifts |
| Tier 2 · Quality & Accuracy Guardrails (Damage Prevention & Sentiment) | |||||||
| 4 | Mention SentimentNet Sentiment Score (-100 to +100) | Real-Time | > +65 | +74 | +6 pts | POSITIVE | Escalate negative snippets to product & PR |
| 5 | Factual Accuracy RateVerified pricing & feature claims | Bi-Weekly | > 95.0% | 96.8% | +1.8% | VERIFIED | Publish schema updates for wrong pricing claims |
| 6 | Crawler ReachabilityHTTP 200 success on key endpoints | Daily | 100.0% | 99.9% | 0 errors | PROTECTED | Audit robots.txt and sitemaps daily |
| Tier 3 · Lagging Outcomes & Board Metrics (Pipeline & Closed ARR) | |||||||
| 7 | Branded Search VolumeGSC impressions (AI discovery lift) | Monthly | +10.0% YoY | 48,500/mo | +14.3% | HIGH LIFT | Attribute top-of-funnel lift to AI engines |
| 8 | AI Referral SessionsGA4 custom AI channel grouping | Monthly | > 3,000 | 3,840 | +22.1% | SURPASSED | Improve landing page paths for AI visitors |
| 9 | AI-Sourced SignupsFirst-touch origin cookie join | Monthly | > 140 | 162 signups | +15.7% | EXCEEDING | Calibrate CAC against blended organic channels |
| 10 | AI-Influenced PipelineCRM opportunities with AI touchpoint | Monthly | > $1.20M | $1.42M ARR | +18.3% | ABOVE PLAN | Feed AI query intent data to enterprise AE teams |
| 11 | Recognized Net-New ARRStripe closed-won AI revenue | Monthly | > 5.0% total | $285k (7.4%) | +1.9% share | BOARD VALIDATED | Reallocate budget: expand 2027 AEO investment |
| Read the scorecard as a build protocol: Only keep rows where the underlying data source is instrumented. Next review: end of month | |||||||
Download the scorecard as HTML →
The tiers follow Kevin Indig's AI visibility ladder, published in Kyle Poyar's Growth Unhinged in July 2026. Indig uses it with Airbnb, Asana and Xero, and credits Andy Grove's High Output Management for the idea of a ladder of leading and lagging indicators. When revenue lags, it tells you whether you are on track long before the revenue arrives. Three setup rules show how to measure AI visibility in a way you can trust:
- Freeze 20 to 50 high-intent prompts across personas, use cases and buying stages for at least four weeks, so you measure real change rather than prompt drift.
- Log every run: prompt, model, location, answer, cited URLs, brands mentioned and shortlist position. That table is the raw material for every rung.
- Run two clocks. Weekly the team checks signal quality. Monthly the CMO checks allocation.
Indig also names three traps: vanity metrics treated as the destination, false precision from decimals on a number that re-rolls monthly, and mixing leading indicators with outcomes without a model that connects them. He advises reporting movement across the ladder rather than a single AEO score. The quote near the top of this page gives his view on last-click measurement.
Every row needs a decision and an instrument
A number that triggers no decision, or that you have no way to move, is the first trap on that list. Read this table as a build list. Any row where the instrument is not yet true should stay off the scorecard. Three real, actionable rows beat twelve empty scores.
| # | Metric | The decision it triggers | The instrument required |
|---|---|---|---|
| Leading indicators | |||
| 1 | Bot crawls (search crawlers only) | Unblock search crawlers if the count drops sharply | Cloudflare Radar or server logs filtered to AI user agents |
| 2 | Citation share on frozen prompts | Find which reviews and comparison sites took citations | Automated prompt sampling across engines, frozen prompt set |
| 3 | Share of voice vs named rivals | Move content toward topics where rivals lead | Competitor co-mention tracking in the same answers |
| Quality guardrails | |||
| 4 | Sentiment of the mention | Flag negative sentiment to product and PR | Sentiment classification on answer extracts |
| 5 | Factual accuracy | Publish corrections for wrong claims | Answers compared against pricing and feature specs |
| 6 | Crawler reachability | Fix robots.txt, paywalls or rendering that block bots | Daily checks on robots.txt and sitemap URLs for AI bots |
| Lagging indicators | |||
| 7 | Branded search volume | Credit AI discovery when lift tracks citation share | Search Console branded queries with lag analysis |
| 8 | AI referral sessions | Study landing paths for high-intent visitors | GA4 custom channel group for AI referrers |
| 9 | AI-sourced signups | Compare AI signup rate with organic | First-party origin cookie in a hidden form field |
| 10 | AI-sourced paid conversions | Scale or trim AEO budget on CAC | CRM deal join with a read-only origin field |
| 11 | New ARR from AI-sourced | Report ROI to the board and adjust budget | Billing data joined to CRM closed-won deals |
| Source: ThriveStack citedby Research, after Kevin Indig's ladder | |||
The way most teams go about it right now is by measuring pipeline from referral clicks. But that's like valuing a Super Bowl ad by QR-code scans. It under-attributes the impact.
How to measure the impact of AI search the right way, 15 Jul 2026
Which marketing KPIs fit the new AI customer journey?
Six groups of marketing KPIs fit the new journey, one per stage: visibility, quality, traffic, engagement, pipeline and revenue. The buyer no longer starts on a search results page. They ask an assistant, read a shortlist and form a view before they reach your site. A report that opens with referral sessions has skipped two stages.
The old funnel let you watch every step. The new one does not. The first two stages happen inside someone else's app, which is why visible AI clicks are about 1.08% of web traffic while the influence behind them is much larger.
The evaluation moved too, and that is the part most teams miss. In the search era a buyer opened your page, a rival's page and a review site, then compared them across tabs. Your pricing page did the work. Now the assistant does it. It reads all three sources, weighs them and hands back one shortlist with the comparison already made.
So the pages you built to persuade are read by a model before a human sees them, and often instead of a human seeing them. You cannot rank in an answer. You are either in it or you are out, so the first metric is how often you appear.
| # | What changed | Search era | AI answer era |
|---|---|---|---|
| 1 | What the buyer types | Keywords. Two or three words, stripped of context | A prompt. A full question with their situation in it |
| 2 | What they get back | Ten links to choose from | One answer with a short shortlist |
| 3 | Where they compare | Across several open tabs, after the click | Inside the answer, before any click |
| 4 | How many options | As many as they open tabs for | Three to five, and the first is picked most |
| 5 | What you can see | The query, the rank, the click | No query, no rank, and often no click |
| 6 | Where they evaluate | Across your site, review sites and rival tabs | Inside the answer, before any site is opened |
| 7 | What you optimize | Ranking for a keyword | Being cited in an answer |
| Source: ThriveStack citedby Research | |||
| # | Journey stage | What the buyer does | The metrics that see it |
|---|---|---|---|
| 1 | Discovery | Asks an assistant, reads a shortlist | Visibility. Citation share, brand mention rate, competitor co-mention |
| 2 | Shortlist | Forms a view from how you are described | Quality. Sentiment, factual accuracy, crawler reachability |
| 3 | Visit | Clicks through, or searches your brand later | Traffic. AI referral sessions, branded search, landing page mix |
| 4 | Evaluation | Explores the product and the pricing page | Engagement. Bounce, duration, pages per session |
| 5 | Intent | Signs up, requests a demo, enters pipeline | Pipeline. AI-sourced opportunities, win rate, cycle length |
| 6 | Purchase | Converts to a paid plan | Revenue. Paid conversions, new ARR, CAC payback |
| Source: ThriveStack citedby Research | |||
Why does AI share of voice replace the old rank report?
AI share of voice replaces the rank report because position in an answer keeps changing. SparkToro and Gumshoe.ai ran 2,961 tests with 600 volunteers. ChatGPT and Google's AI gave the same brand list twice under 1 time in 100. The same list in the same order came closer to 1 in 1,000.
This section explains scorecard rows 1 to 6. Report share of voice and share of model across many runs instead. Rand Fishkin's verdict on the study is blunt: "any tool that gives a 'ranking position in AI' is full of baloney." Gumshoe co-produced the research, and SparkToro flags that possible conflict in the post.
Links are also a poor proxy. BrightEdge measured about 2.4 ChatGPT brand mentions per prompt against 0.74 citations, or 3.2 times more. A brand named without a link can still move demand.
Quality rows sit beside volume because being named badly is worse than being left out. Check sentiment and factual accuracy on every mention, and confirm AI bots can reach your pages. Compare your citation share with peers in our 2026 AI visibility benchmarks.
| # | Metric | How to calculate | What it tells you |
|---|---|---|---|
| 1 | Citation share | Prompts where you are cited ÷ prompts tracked | Presence with a linkable source |
| 2 | Brand mention rate | Answers naming you ÷ answers generated | Influence beyond citations |
| 3 | Prompt coverage | Prompts where you appear ÷ mapped buyer prompts | Breadth across the buying journey |
| 4 | Average citation position | Mean rank within the answer shortlist | Directional only. Never report it as a rank |
| 5 | Competitor co-mention | Your answers also naming rival X ÷ your answers | The comparison set engines placed you in |
| 6 | Model variance | Citation share by engine, plus cross-engine overlap | Writesonic found only 3.8% of sources shared by all four engines it tested |
| Source: ThriveStack citedby Research. Writesonic, July 2026 | |||
Why does AI referral traffic miss most of your visits?
AI referral traffic misses most visits because buyers read the answer, then come back another way. Similarweb found 55.9% of AI-influenced visits arrive through search, against about 40% of standard visits, and over half arrive through a branded query. Its study covers ChatGPT recommendations only. Every number here is a floor, so say so.
This section explains scorecard rows 7 and 8. The figure shows the split. Above the line, analytics sees the referrer, the landing page and the same-session conversion. Below it, branded searches, visits from another device and decisions made in chat all land in other channels or nowhere.
Search Console now helps a little. In June 2026 Google announced dedicated reports for impressions in AI Overviews and AI Mode, and they reached all sites by the end of August. Those are impressions, so they belong at the discovery stage.
Branded search is where the influence lands. Similarweb found brands ChatGPT recommended were 2.5 times more likely to get a site visit within seven days. When citation share rises and branded search follows within a sales cycle, you have a story you can defend. When nothing follows, the citations are likely low quality.
| # | Metric | Source and calculation | Benchmark |
|---|---|---|---|
| 1 | AI referral sessions | GA4 sessions matching the AI channel regex | The measurable click floor |
| 2 | AI referral share | AI sessions ÷ total sessions | 1.08% across 1,215 enterprise domains. IT 2.80%, Utilities 0.35% |
| 3 | Growth rate | Month over month change in AI sessions | About 1% month over month on the Conductor panel |
| 4 | Landing page mix | AI sessions grouped by landing page | Check product and comparison pages against blog |
| 5 | Engine mix | AI sessions by referrer host | ChatGPT is 87.4% of AI referrals |
| Source: Conductor 2026 AEO / GEO Benchmarks Report, May to Sep 2025 | |||
Never report crawler hits as traffic. Cloudflare measured about 70,900 crawls per referred visitor for Anthropic in June 2025, so mixing bots into this stage inflates it by orders of magnitude. Cloudflare also notes that native apps send no referrer, so its ratios may run high. For setup, see our guide to GA4 AI traffic.
What does the evidence say about AI search conversion rate?
AI search traffic usually converts better than other channels, and the largest datasets agree. Adobe's retail data covers over 1 trillion visits and shows AI referrals converting 90% higher in September 2026, the highest level it has observed. Microsoft Clarity found LLM sign-ups at 1.66% against 0.15% for search. One 54-site study found no significant gap.
This section explains scorecard row 9, AI-sourced signups. The right multiple depends on what counts as a conversion, so compare like with like. Retail purchases and sign-ups are different events, and the chart keeps them apart. Monthly figures also swing. Adobe reported 60% higher in July 2026, so read any single month with care. Adobe compares against non-AI traffic, while Clarity's sample is publisher and news sites.
1.66% vs 0.15% for search
1.34% vs 0.55% for search
Plan on about 1.5× for retail purchases and 4× or more for sign-ups, then measure your own, because single-site multiples like 23× are not a planning number.
How does pipeline attribution win more marketing budget?
Pipeline attribution wins budget because it puts a dollar figure on a channel the CFO cannot otherwise see. Every earlier stage points at this one. Report AI-sourced and AI-influenced pipeline created, win rate, average deal size and cycle length, each split by first-touch source, and keep the origin field read only.
This section explains scorecard row 10, AI-influenced pipeline. The figure shows the chain that has to hold. The second handoff is where most setups fail. If the origin never reaches the CRM, AI influence gets credited to Direct or Organic Search, and the number you need for the budget meeting is gone. Read our guide to revenue attribution for the join itself.
| # | Metric | Calculation | Why it matters |
|---|---|---|---|
| 1 | AI-sourced pipeline | Sum of opportunity value where the AI origin flag is set | The headline number for a CRO |
| 2 | AI-influenced pipeline | Adds opportunities where self-reported source names an AI tool | Recovers the no-click majority |
| 3 | Win rate by source | Closed-won ÷ opportunities, split by first touch | Reveals quality that lead counts hide |
| 4 | Average deal size | Closed-won value ÷ deals, split by first touch | Fewer, larger deals can beat higher volume |
| 5 | Sales cycle length | Median days from first touch to close | Pre-qualified buyers should close faster |
| 6 | CAC by source | Channel cost ÷ customers acquired | Makes the ROI question answerable |
| Source: ThriveStack citedby Research | |||
Which revenue attribution metrics count paying customers?
The revenue attribution metrics that count start with a new customer on a paid plan. That is the moment a card is charged for the first time, or a contract closes. A trial start does not count. Nor does an MQL, or pipeline created. Lead with AI-sourced paid conversions, then new ARR from those accounts.
This section explains scorecard row 11, new ARR. The figure, also shown at the top of this page, draws the line. AI search sits at the start of the journey. A signup proves the citation was seen. A first payment proves it was worth something. Every metric below is anchored to the payment.
Both paths need attribution, and they fail in different ways. On the self-serve path the whole chain is observable: AI answer, signup, activation, first charge. It can close in days, so the origin flag rarely has time to get lost.
The sales-led path is harder. The same buyer takes months, several people from the account get involved, and the person who signs the order is almost never the person who asked the assistant. Roll touches up to the account before assigning credit. Otherwise the channel that created the demand looks like it contributed nothing.
| # | Metric | Calculation | Caveat |
|---|---|---|---|
| 1 | AI-sourced paid conversions | New customers on a paid plan carrying an AI origin flag | The primary number. Count customers rather than signups |
| 2 | Signup to paid rate | AI-sourced paid conversions ÷ AI-sourced signups | Compare against your blended rate rather than an industry figure |
| 3 | Time to first payment | Median days from AI-sourced signup to first charge | Pre-qualified buyers should convert faster |
| 4 | New ARR from AI-sourced customers | Sum of first-year contract value on those accounts | State the attribution model behind it |
| 5 | Revenue per AI session | AI-sourced new ARR ÷ AI referral sessions | Denominator is a floor, so the figure runs high |
| 6 | Revenue per citation | AI-sourced new ARR ÷ citations in period | Requires a closed loop. Rare in practice |
| 7 | CAC payback on AI-sourced | Program cost ÷ new ARR from AI-sourced customers | Treat vendor-published ROI figures with care |
| 8 | Incremental lift | Paid conversions against a matched holdout | The only causal measure, and rarely feasible |
| Source: ThriveStack citedby Research | |||
At current volumes most companies cannot produce a statistically significant per-citation revenue figure. Say that in the first meeting rather than being forced into it later. What you can defend is the AI-sourced customer count and the direction of the trend, shown in your attribution reporting with the method named beside each figure.
What should you ask AI visibility tools before you buy one?
Ask AI visibility tools three things: show me a closed-won deal, tell me my sample size, and tell me if you report rank. Most of the category measures presence in answers and stops at the citation. These questions separate a dashboard from an attribution system, and each one has a clear wrong answer.
- Can you show me a closed-won deal and the AI answer that started it? Few vendors can answer this.
- How many prompts, runs and engines does my tier cover? Many tools price on tracked prompts. On October 30, 2026, Otterly's plans started at 15 prompts and Peec's at 50. Given the variance above, a small set run once tells you little.
- Do you report ranking position? If yes, they have not absorbed the research on how much answers vary.
Search demand is climbing too. AI visibility tools went from 1,000 monthly US searches in October 2025 to 1,300 in September 2026. For named options, see our best AEO tool scorecard and market landscape of 50 vendors.
How does ThriveStack citedby differ from other AEO tools?
ThriveStack citedby differs from other AEO tools because it carries the AI origin through to paying customers. This metric set is only useful if the last stage is filled in. Most tools stop at the citation. We sample answers across about ten engines, track which pages get cited, then join that origin to CRM and billing data.
ThriveStack citedby is the AI visibility and answer engine optimization (AEO) product inside ThriveStack. The loop it runs is analyze, fix, attribute. It reports each stage above:
- Citation share, mention rate and co-mention on a frozen prompt set across about ten engines, reported as movement with confidence intervals.
- Sentiment and factual accuracy of each mention, so quality sits beside volume in one view.
- AI referral capture covering the engines GA4's native channel omits, including Perplexity and Claude.
- Branded search and direct correlated against citation share with lag, which recovers the influence that lands elsewhere.
- AI-sourced pipeline and closed-won revenue through CRM and billing integration, which is the stage that renews budget.
Build this report from your own data
ThriveStack citedby tracks citation share, mention quality, AI referrals and AI-sourced pipeline in one view, connected to your CRM and billing data.
Start free trial →AI visibility metrics: FAQ
What AI visibility metrics should a CMO track?
Six groups, one per stage of the new customer journey. Discovery: citation share, mention rate, co-mention. Shortlist: sentiment, factual accuracy, crawler reachability. Visit: AI referral sessions and branded search. Evaluation: bounce, duration, conversion rate. Intent: AI-sourced opportunities, win rate, deal size. Purchase: paid conversions and new ARR. Report movement against named competitors rather than absolute scores.
How do you measure AI visibility?
Freeze 20 to 50 high-intent prompts, run them repeatedly across the engines your buyers use, and log every answer, cited URL and brand named. Report citation share and mention rate against named rivals. Then connect those signals to branded search, AI-sourced pipeline and paid customers so the numbers lead to a budget decision.
Should AI ranking position be on the report?
No. SparkToro and Gumshoe.ai found under a 1 in 100 chance of ChatGPT or Google's AI returning the same brand list twice across 2,961 runs, and roughly 1 in 1,000 for the same order. Position in AI answers moves too much to measure.
What is the difference between citation share and mention rate?
Citation share counts answers where you appear with a linkable source. Mention rate counts answers naming you at all, with or without a link. BrightEdge measured about 2.4 ChatGPT mentions per prompt against 0.74 citations, so tracking citations alone misses most of the influence.
When does AI visibility justify more budget?
Our rule of thumb: when AI-sourced and AI-influenced pipeline exceeds roughly 5% of new pipeline, or when branded search rises in step with citation share over a full sales cycle. Below that, hold at maintenance spend rather than arguing from leading indicators alone.
Which AI visibility metric best predicts revenue?
Branded search lift is the best early signal we have seen. Similarweb found brands ChatGPT recommended were 2.5 times more likely to receive a visit within seven days, and 55.9% of those visits arrived through search. Ahrefs put branded web mentions against AI Overview visibility at 0.664, while backlinks and ad cost sat near 0.22.
What goes on a CMO dashboard for AI visibility?
Three rungs, following Kevin Indig's AI visibility ladder in Growth Unhinged, July 2026. Leading indicators: bot crawls, citation share, share of voice. Quality guardrails: sentiment, factual accuracy, crawler reachability. Lagging indicators: branded search, AI referrals, AI-sourced signups and paid conversions. Report movement across the rungs rather than a single AEO score.
What counts as revenue in AI search attribution?
A new customer converted to a paid plan, meaning the first charge or a closed contract rather than a trial start, an MQL or pipeline created. AI search operates at the discovery end of the journey, so a signup only proves the citation was seen. The first payment proves it was worth something.
What should you ask an AI visibility vendor?
Whether they can show a closed-won deal and the AI answer that started it, how many prompts, runs and engines your tier covers, and whether they report ranking position. The last one tests whether they have read the research on how much answers vary.
Sources
- SparkToro and Gumshoe.ai, “AIs are highly inconsistent when recommending brands or products”. January 2026. 600 volunteers, 2,961 runs.
- Similarweb, “The Downstream Impact of AI Visibility”. June 2026. ChatGPT recommendations in Finance, Travel and Beauty.
- Amsive, “Does LLM Traffic Convert Better Than Organic?”. September 2025. 54 sites, six months of GA4 data.
- Ahrefs, “Does AI Search Traffic Convert Better Than Traditional Search?”. June 2025. Ahrefs' own site.
- Conductor, 2026 AEO / GEO Benchmarks Report. Traffic findings from 1,215 enterprise domains, May to September 2025.
- BrightEdge, “ChatGPT brand mentions vs citations”. August 2025.
- Ahrefs, AI Overview brand correlation study. May 2025, updated April 2026.
- Ahrefs, AI brand visibility correlations. December 2025, updated August 2026.
- Writesonic, AI citation source overlap study. July 2026.
- Adobe, AI traffic to US retail sites, July 2026 data. August 2026.
- Adobe, AI traffic to US retail sites, March 2026 data. April 2026.
- Adobe, generative AI shopping traffic, April 2025 comparison. August 2025.
- Adobe, AI traffic and revenue per visit, September 2026 data. October 2026.
- Microsoft Clarity, AI traffic conversion study. November 2025. 1,277 publisher and news domains.
- Semrush, AI search SEO traffic study. July 2025.
- Seer Interactive, how traffic from ChatGPT converts. June 2025. One client.
- Google Search Central, “Introducing Search Generative AI performance reports in Search Console”. 3 June 2026.
- Cloudflare, “The crawl before the fall of referrals”. 1 July 2025.
- Kevin Indig, “How to measure the impact of AI search the right way”. Growth Unhinged (Kyle Poyar), 15 July 2026.
