Google’s own Conversational Analytics API documentation warns that Gemini for Google Cloud products can generate output that seems plausible but is factually incorrect (Google Cloud, 2026). On a WooCommerce store missing 20-35% of conversions to ad blockers and browser restrictions, that warning is doing a lot of work.
GA4’s Analytics Advisor — the Gemini-powered conversational interface inside Google Analytics — is now the default AI-recommended path for non-technical GA4 users. It is fluent, confident, and well-phrased. It is also reading whichever data GA4 can see, which on a typical SMB WooCommerce store is a 15-50% underreport of actual revenue (Cardinal Path / Seresa analysis, 2026). The AI is not lying. The data layer underneath it is.
Why Analytics Advisor’s Confident Answers Inherit GA4’s Data Gap
Analytics Advisor was announced at Google Marketing Live in May 2025 and has been rolling out progressively through 2026. Looker Conversational Analytics reached general availability in November 2025 (Google Cloud Blog, 2025), and the same Gemini engine now sits inside GA4’s standard interface as a recommended workflow for asking strategic questions in plain English.
What it reads matters. Inside the GA4 UI, Analytics Advisor queries the GA4 reporting layer — sessions, events, conversions, audiences — generated by Gemini SQL against the GA4 backend. It does not natively touch your BigQuery export, your WooCommerce orders table, or any source outside GA4. Whatever data loss happens upstream of GA4 is invisible to the AI.
And there are four big losses upstream, each documented, each silent.
The Four Silent Gaps Analytics Advisor Inherits
Gap 1: 31.5% of Your Visitors Are Blocked Before GA4 Sees Them
31.5% of global internet users run ad blockers that block GA4 client-side tracking entirely (Statista, 2024). The gtag.js script never executes, the page_view never fires, the conversion never reaches the GA4 collection endpoint. From GA4’s perspective, those users do not exist.
Ask Analytics Advisor “which channel drives the most conversions” and the answer is computed over the two-thirds of users who do not block tracking. If your ad-blocker rate skews by channel — and it usually does, with technical and EU audiences blocking more — the AI’s channel ranking is structurally biased before Gemini writes a single SQL query.
Gap 2: Safari ITP Quietly Resets Attribution Windows
Safari’s Intelligent Tracking Prevention caps JavaScript-set first-party cookies at 7 days for cross-site-classified domains. On any consideration cycle longer than a week — which is most ecommerce — Safari users are recorded as new visitors on every revisit, and their attribution paths are truncated to the last 7 days.
The Analytics Advisor question “what does my customer journey look like” returns a journey graph that is artificially short for roughly a third of your traffic. The AI does not flag this. It cannot see what it is not given.
Gap 3: Data-Driven Attribution Silently Falls Back to Last-Click
GA4 data-driven attribution requires at least 400 conversions per month to activate; below that threshold GA4 silently falls back to last-click (Google Analytics Help, 2025). Most WooCommerce SMB stores never cross 400 monthly conversions per property — the threshold is a hard cliff, and the fallback ships without a banner, log entry, or status line in the AI’s context.
This is the part that bites. Ask Analytics Advisor for “the most efficient channel” and the answer is computed against last-click on a sub-DDA-threshold property — but the AI’s phrasing implies a multi-touch model. We covered this fallback in detail in Your WooCommerce Campaign Report Shows Last Click — Not Who Made the Sale.
Gap 4: Predictive Metrics Require Volume You Don’t Have
GA4 predictive metrics require 1,000+ users in each behavioural cohort over 7-28 days to activate (Anomaly AI, 2026). Purchase probability, churn probability, predicted revenue — none of them switch on for most SMB stores.
The Analytics Advisor will still answer questions that imply the metrics exist. Ask it “who is most likely to convert next month” and Gemini will compose a confident reply built from whatever proxies are available, presenting the synthesis as if the predictive layer were active. It is not.
The “Ask It Something You Already Know” Test
Run this five-minute test today.
Pull last month’s WooCommerce orders report — the one your finance team uses, generated server-side from the WordPress database. Note the total revenue, total order count, and channel breakdown if you have it.
Now open Analytics Advisor and ask: “What was last month’s total revenue and how did it compare to the previous month?”
Compare the AI’s answer to the WooCommerce truth. If the gap is under 5%, your tracking chain is healthy. If the gap is 15-35%, you have just confirmed which slice of strategic advice the AI is rounding off. The AI is not lying. The data layer underneath it is.
This is the same structural reason dashboards struggle to query GA4 directly — we wrote about it in Your Looker Studio Dashboard Takes 45 Minutes to Load Because It’s Asking GA4 Questions GA4 Can’t Answer. Analytics Advisor inherits the same constraint, just with prettier prose.
How to Give Gemini Complete Data
The fix is not to abandon the AI. The AI is good at what it does — translating natural-language questions into structured queries and summarising results in plain English. The fix is to feed it complete data instead of browser-pixel-filtered data.
Two moves.
First, export GA4 events to BigQuery. The export is free at SMB volumes, gives you the raw event-level data, and unlocks the Conversational Analytics API which can query BigQuery datasets directly with Gemini-generated SQL.
Second, capture WooCommerce orders server-side and route them into the same BigQuery dataset. Transmute Engine™ is a dedicated Node.js server that runs first-party on your subdomain (e.g., data.yourstore.com); the inPIPE WordPress plugin captures WooCommerce hooks and sends events via API to Transmute Engine, which then formats and routes them simultaneously to GA4, Facebook CAPI, BigQuery, and Google Ads — bypassing every browser-side gap above. The same Gemini, asking the same questions, against complete data.
Key Takeaways
- Google’s own docs warn Gemini can sound plausible while being wrong. The Conversational Analytics API documentation says it explicitly. On incomplete WooCommerce data, that warning is the headline.
- GA4 underreports WooCommerce revenue by 15-50% from ad blockers (31.5% of users), Safari ITP, and consent-rejected traffic combined.
- Data-driven attribution silently falls back to last-click below 400 monthly conversions. The AI inherits the fallback without disclosing it.
- Predictive metrics need 1,000+ users per cohort to activate — most SMB stores never qualify, but Analytics Advisor will answer as if they do.
- Run the “ask it something you already know” test. Compare Analytics Advisor’s revenue answer to your WooCommerce dashboard. The gap is the slice the AI is rounding off.
- The fix is BigQuery + server-side capture. Same Gemini, complete data underneath.
Frequently Asked Questions
Analytics Advisor reads your GA4 property’s standard reports — sessions, conversions, events, audiences — generated by Gemini queries against the GA4 backend. It does not natively query your BigQuery export, your WooCommerce orders table, or any source outside GA4 unless you’ve connected it via the Conversational Analytics API. Whatever browser-side data loss exists upstream of GA4 is invisible to the AI, which has no signal that the dataset is incomplete.
Not by default. The standard Analytics Advisor inside the GA4 UI reads only GA4’s reporting layer. Google’s separate Conversational Analytics API can query BigQuery datasets and Looker semantic models with Gemini-generated SQL, but you have to wire it up explicitly. Until you do, the AI is answering from the same browser-pixel-filtered data your dashboard shows — not from your actual order data.
Trust the WooCommerce dashboard. WooCommerce records orders server-side at the database level — no browser, no ad blocker, no cookie. If GA4 disagrees, the gap is your tracking chain. Until the AI is reading complete data, its strategic advice is rounding off whichever 20-35% it can’t see, and Google’s own docs already warn the output can sound plausible while being factually incorrect.
No. GA4 data-driven attribution requires 400+ monthly conversions per property to activate. Below that, it silently switches to last-click without a banner, log entry, or warning to the AI layer. Analytics Advisor will confidently describe channel performance using last-click numbers while implicitly framing them as DDA. The fallback is a documented Google Analytics behaviour — the AI just doesn’t surface it.
Two moves. First, export GA4 events to BigQuery (free at SMB volumes) so you have the raw event-level data. Second, capture WooCommerce orders server-side with a first-party tracking server like Transmute Engine and route the same orders into the same BigQuery dataset. Then the Conversational Analytics API queries the merged dataset — the AI is the same Gemini, asking the same questions, but the data underneath finally matches your store.
Run the test today. Compare what GA4’s Gemini tells you against what WooCommerce already knows — then close the gap at the data layer, not the prompt. Start at seresa.io.
