AI and search visibility becomes pipeline through four paths: a buyer clicks a citation, searches your brand after reading an answer, returns directly later, or tells sales an assistant suggested you. Only the first path is tracked by default. Measuring pipeline means joining GA4's AI Assistant channel, Search Console, self-reported attribution and CRM opportunity data, then reporting sourced and influenced pipeline alongside visibility.
This is the final part of our series on how search and AI visibility work. The first six parts explained how pages get retrieved, quoted and recommended. This part explains how to prove any of it produced revenue. For a deeper model, read AI search attribution. For the GA4 build in detail, read measuring AI referral traffic in GA4. For why AI visitors convert differently, read AI referral traffic conversion rate.
The short version
- Visibility An engine names or cites your company in an answer to a buyer's question.
- Exposure The buyer sees it, which Search Console and Bing count as impressions or citations.
- Visit The buyer clicks through, searches your brand, or comes back directly later.
- Conversion The buyer completes a key event such as a demo request, captured with its source.
- Opportunity Sales qualifies the lead and the CRM records sourced and influenced attribution.
- Revenue Closed deals are traced back to the paths that started or shaped them.
The four paths from an AI answer to a sales conversation
B2B buyers rarely click a citation, fill in a form and buy. They read an answer, note two or three vendor names, and come back later, often on another device, often after talking to colleagues. Each path leaves a different trace.
| Path | What the buyer does | Where it shows up | Tracked by default | How to capture it |
|---|---|---|---|---|
| 1. Citation click | Clicks a link inside ChatGPT, Perplexity, Gemini, Copilot or Claude | GA4 AI Assistant channel | Yes, when a referrer is passed | Confirm the channel; pass source into CRM with the form |
| 2. Google AI feature click | Clicks a link in an AI Overview or AI Mode | Search Console Web search type; organic search in GA4 | Counted, not separated in GA4 | Search Console page and query exports; generative AI performance report |
| 3. Brand search | Reads an answer, then searches your company name | Branded queries in Search Console; organic search in GA4 | No link to the AI answer | Branded query trend; self-reported attribution |
| 4. Direct or dark visit | Types your URL, uses an app that strips the referrer, or forwards your name to a colleague | Direct in GA4, or nowhere | No | Self-reported attribution; sales discovery question |
| No visit at all | Reads the answer and never clicks | Prompt benchmarks; Bing citation counts | No | Visibility tracking, reported as a leading indicator |
The table explains why AI search looks smaller in analytics than it is in the market. Only one row is fully attributed by default. Our piece on zero-click B2B marketing covers the last row in depth.
Setting up the measurement chain, step by step
Step 1: Confirm the AI Assistant channel in GA4
Google's default channel group documentation lists an AI Assistant channel for users arriving from sources such as ChatGPT, Gemini, Deepseek, Copilot or Grok. Per that documentation, when the referrer matches Google's list of AI assistants, GA4 sets the medium to ai-assistant and the campaign to (ai-assistant). The channel explicitly excludes Google's AI Overviews and AI Mode.
Open Reports, then Acquisition, then Traffic acquisition, and set the primary dimension to Session default channel group. Then switch to Session source and check which assistant domains appear, since the list of recognized sources is Google's and may not match every assistant your buyers use.
Step 2: Add a custom channel group if you need control
If you want your own definition, for example to report Perplexity and Claude separately or to catch sources the default list misses, GA4 supports custom channel groups. Go to Admin, then Channel groups, and create a new group or copy the default. Add a channel with a condition on Session source using "matches regex":
.*(chatgpt\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com).*Order matters, because traffic is assigned to the first channel whose conditions it matches, so place the AI channel above Referral. Google says custom channel groups apply to reports retroactively. Standard properties can have two custom groups and GA4 360 properties five, each with up to 50 channels. Validate the expression against the source values GA4 has actually recorded for your property before relying on it.
Step 3: Mark the events that matter as key events
GA4 calls conversions key events. In Admin, then Events, mark the events that signal sales intent, such as demo requests, contact sales submissions and trial signups, by clicking the star. Keep the list short. A newsletter signup is a key event for a media company and noise for a B2B pipeline report.
Step 4: Carry the source into the CRM
GA4 knows about sessions. Your CRM knows about opportunities. Join them at the form. Capture the session source, medium, landing page and referrer in hidden fields when a form is submitted, and store both the first-touch and the latest-touch values on the contact record. Without this step, an AI-referred demo request becomes an anonymous "website" lead the moment it leaves GA4.
Step 5: Ask buyers directly
Self-reported attribution recovers paths 3 and 4. Add a required, open-text "How did you hear about us?" field to demo and contact forms, and classify answers monthly into categories that include "AI assistant". Add one question to sales discovery calls, "What did you use to research vendors?", and log the answer in a CRM field. Buyers who say "ChatGPT suggested you" are telling you something no referrer header will.
Step 6: Define sourced and influenced pipeline before you count
Write the definitions down and get sales leadership to agree before the first report.
- AI-sourced pipeline: opportunities whose first recorded touch is an AI Assistant session, or whose self-reported source is an AI assistant.
- AI-influenced pipeline: opportunities where any associated contact had an AI Assistant session, named an AI assistant in self-reported attribution, or told sales they used one, within a set lookback window before the opportunity was created. Ninety days is a reasonable starting window for most B2B cycles; choose yours and keep it fixed.
Influenced pipeline will always be larger than sourced pipeline and will overlap with other channels. That is expected. Report both, and never add influenced pipeline across channels as if it were additive.
Step 7: Read GA4 attribution for what it is
GA4's attribution documentation lists three models: data-driven, paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based models were removed in November 2023. One detail matters for AI traffic: all models exclude direct visits from receiving credit unless the whole path is direct. A buyer who arrives from ChatGPT and later returns directly to convert gives the credit to the AI touch, not to Direct, within the limits of GA4's own tracking.
Those limits are the real constraint. B2B deals involve several people, devices and months. GA4 is good for comparing channel behavior. The CRM, with sourced and influenced fields, is where pipeline is counted.
Exposure metrics that sit upstream of traffic
Two free sources show AI exposure before anyone clicks. In Google Search Console, AI Overview and AI Mode activity is included in the Performance report under the Web search type; Google counts an AI Overview as a single position shared by all its links, and its generative AI guide points to the generative AI performance report for visibility in those features. In Bing Webmaster Tools, AI Performance shows total citations, average cited pages, sampled grounding queries and page-level citation activity across Copilot and Bing AI summaries. Neither covers ChatGPT, Claude or Perplexity, which is why a prompt benchmark still sits at the top of the chain; see how to benchmark AI visibility.
What to report to the board
Boards need to know three things: whether buyers see you, whether that turns into pipeline, and what you will do next. A one-page report in four layers answers all three.
| Layer | Metric | Source | Cadence | How to present it |
|---|---|---|---|---|
| Visibility | Mention rate, citation rate and share of voice on a fixed buyer prompt set, against three named competitors | Prompt benchmark | Monthly, reported quarterly | Trend line per engine; note prompt set changes |
| Exposure | Impressions and clicks for priority pages; Bing AI citations | Search Console, Bing Webmaster Tools | Monthly | Quarter over quarter change, not daily noise |
| Engagement | AI Assistant sessions, key event rate compared with other channels | GA4 | Monthly | Labeled as a floor, with the Direct caveat |
| Pipeline | AI-sourced and AI-influenced opportunities, value, win rate and cycle length compared with other sources | CRM | Quarterly | Counts and value side by side, with definitions in a footnote |
| Decisions | What changed, what it cost, what you will do next quarter | Marketing plan | Quarterly | Three bullets tied to the layers above |
Three rules keep the report credible. Show trends over at least two quarters before claiming an effect. Never present influenced pipeline as caused by AI search. And put the visibility layer first, because it moves before revenue does. If the board wants a financial frame, our AEO business case and cost per AI citation guides show how to model it.
Five measurement mistakes that undercount or overclaim
- Treating the AI Assistant channel as the total. It records only referred clicks from recognized assistants. Paths 2, 3 and 4 sit in other channels or nowhere.
- Tagging your own links in ways that hide the real source. UTM parameters on links you place in documentation, directories or partner listings replace the referrer GA4 would otherwise use. Keep campaign tagging for campaigns you actually run.
- Counting Google AI features twice. AI Overview and AI Mode clicks already sit inside organic search. Report them from Search Console, not as an extra line added on top of GA4 totals.
- Changing the prompt set every month. Visibility trends only mean something if the prompts stay fixed. Add new prompts as a separate cohort.
- Reporting before sales agrees on definitions. An influenced pipeline figure that sales leadership never approved will be challenged in the first board meeting, and the whole channel loses credibility with it.
One sanity check is worth running every quarter: compare the share of opportunities with a self-reported AI answer against the share with an AI Assistant session. A large gap is normal. It shows how much of AI search's effect happens off the tracked path, which is exactly what the board needs to understand before judging the channel on GA4 numbers alone.
What you control: the pipeline measurement checklist
- The AI Assistant channel is visible in your standard GA4 acquisition report.
- A custom AI channel, if used, sits above Referral and is validated against recorded sources.
- Key events are limited to actions that signal sales intent.
- Forms pass source, medium, landing page and referrer into first-touch and latest-touch CRM fields.
- Demo and contact forms include a required open-text "How did you hear about us?" field.
- Sales logs a research-method answer on every discovery call.
- Sourced and influenced definitions, including the lookback window, are written down and agreed with sales.
- Search Console and Bing Webmaster Tools are exported monthly for priority pages.
- A fixed buyer prompt set is benchmarked monthly against named competitors.
- The board report uses the same four layers every quarter.
Illustrative example: one quarter of AI pipeline reporting
Illustrative example. The company and every number below are hypothetical, chosen to show the arithmetic. They are not benchmarks or a client result.
A procurement software company with a sales-led motion closes its first full quarter of AI measurement. The CRM shows 40 new opportunities. The marketing operations lead builds the report in this order:
- Sourced from GA4 data. 3 opportunities have an AI Assistant session as the first recorded touch on the primary contact.
- Sourced from self-reporting. 4 more have "ChatGPT" or "Perplexity" as the answer to "How did you hear about us?", with no AI session on record, most likely paths 3 and 4. AI-sourced total: 7.
- Influenced. Another 6 opportunities were sourced by events or outbound, but a contact on the account had an AI Assistant session within 90 days, or told sales they used an assistant to build a shortlist. AI-influenced total, including sourced: 13.
- Deduplicate. Each opportunity is counted once per category, so an opportunity with both an AI session and a self-reported AI answer counts once.
- Context. The prompt benchmark shows citation rate rising in two of four engines over the quarter, and Search Console shows impressions growing on the pricing and comparison pages that engines cite most.
- Report. The board slide states: 7 of 40 opportunities AI-sourced, 13 of 40 AI-influenced, definitions in the footnote, GA4 figures labeled as a floor, and next quarter's two actions.
Note what the report does not claim. It does not say AI search caused 13 deals. It says AI search was documented in 13 buying journeys, and it shows the visibility trend that explains why. For documented examples of search and content turning into pipeline, see how inbound came to drive 90 to 95 percent of results for Aavenir, and how organic search built $6.4M in pipeline for IQLECT.
Glossary
- AI Assistant channel: a GA4 default channel for sessions referred by recognized AI assistants, excluding Google's AI features.
- AI-influenced pipeline: opportunities where AI search appears anywhere in the documented journey within a lookback window.
- AI-sourced pipeline: opportunities whose first documented touch is an AI assistant.
- Custom channel group: a GA4 set of rules that assigns sessions to channels you define.
- Dark traffic: visits and referrals that arrive without source data, usually recorded as Direct or not at all.
- Data-driven attribution: GA4's model that assigns credit using your property's converting and non-converting paths.
- Key event: GA4's name for an action that matters to the business, formerly called a conversion.
- Lookback window: the period before an opportunity in which touches count toward attribution.
- Self-reported attribution: the buyer's own answer to how they found you.
Where to go from here
That completes the series. If you are starting from scratch, go back to how SEO works and read in order. To put the whole system to work, get the B2B AEO and GEO playbook, which includes the benchmark template and measurement chapter, or request a free audit of your visibility, tracking and pipeline reporting.
