TRACKING SPECIALIST — BARIS ASA · LONDON
Synctag
Start the check → TRACKING SPECIALIST — BARIS ASA · LONDON

READING ROOM › SYMPTOM FILES › SYM-088

SYMPTOM FILE SYM-088 FILED 16 NOV 2026 · REVISED 16 NOV 2026

GA4 events are missing for visitors who decline cookies: what comes back and what doesn't.

ALSO FILED AS
  • “GA4 events missing for visitors who decline cookies”
  • “GA4 not tracking users who reject cookies”
  • “what happens in GA4 when users decline cookies”
  • “GA4 users dropped after cookie banner”
  • “less GA4 data after Consent Mode v2”
  • “GA4 Including estimated user data not showing”
  • “GA4 reporting identity Observed vs Blended consent”

Baris Asa, tracking consultant, London · Last checked: 16 November 2026

Visitors who decline cookies are not tracked as users in GA4. That is expected. In advanced Consent Mode, Google’s tag still sends a cookie-free signal after a “no”, so part of them comes back. GA4 can model their key events (its name for conversions). With enough traffic and Blended reporting identity, it also counts their events and estimates users and sessions. Basic mode estimates no visitors back.
Your set-upWhat GA4 gets back for a “no”
No Consent Mode, banner blocks GA4Nothing
Basic Consent ModeNo visitors; Google’s table marks a general conversion model
Advanced, under the traffic thresholdsNo users or events in reports; key events may be modelled, or credited to Direct
Advanced, thresholds met, Blended identityEvents counted; users and sessions estimated; key events modelled
Any mode, IAB TCF bannerNo modelling in GA4 for denials

Which row are you? Ask whoever set up the banner two questions: does Google’s tag load before the visitor answers (yes is advanced, no is basic), and is the banner running in IAB TCF mode? The browser test is below. What Consent Mode is in the first place: Google Consent Mode v2 explained. Google Ads conversions have their own model and screen: modelled conversions in Google Ads.

A rule of thumb, not a Google figure. Take your decline rate from your consent platform’s dashboard (Cookiebot, OneTrust and similar report accept and reject shares), then compare GA4 purchases with store orders for one closed week.

  • Basic, or TCF: a purchase gap about the size of the decline share is what consent alone produces.
  • Advanced: expect a smaller gap, because modelled key events add some declined purchases back, even under the traffic thresholds when Google is confident enough.
  • Much bigger than the decline share, in any mode: consent doesn’t explain it. Go to when it is a fault.

Decliners don’t always buy at the same rate as accepters, and blockers and browser limits take their own cut, so treat this as a first read. The worked arithmetic across GA4 and Google Ads, store orders against both, is in conversions dropped after Consent Mode v2.

Is it normal that a declined visitor doesn’t show in GA4?

Yes. With analytics_storage denied, Analytics “will not read or write first-party analytics cookies” (Analytics Help: consent mode). No cookie, no client ID, so GA4 can’t join that visitor’s page views into one user.

DebugView won’t show your declined test visit either. Google’s DebugView help says events aren’t visible in debug mode when consent mode is on and the visitor hasn’t consented to analytics cookies (Analytics Help: DebugView). Send these three steps to whoever owns your tags instead:

  1. Fresh incognito window, from the country your customers are in (banners are often set by region). Chrome DevTools, Network tab, filter collect.
  2. Decline the banner. Advanced: requests keep leaving. Basic, or banner-blocked GA4: none.
  3. Accept in a new window. Requests should now leave in every set-up. The full Tag Assistant version: how to check Consent Mode v2 is working.

What does GA4 get back in advanced mode?

Two things, on two different switches. Both come from the cookieless ping, which Google calls “Google Analytics events that contain coarse dimensions which cannot directly identify an individual” (consent mode).

Key events: modelled when Google is confident. “When advertisers use consent mode, key events are modeled for unconsented users.” They appear in the Events, Key events and Attribution reports, and only “when there is high confidence of quality” (Analytics Help: about modeled key events). Google publishes no volume number. On low traffic they may not be reported, or are credited to Direct. That affects which channel gets the conversion, not Direct sessions.

Visitors: estimated once behavioural modelling is on. Conditions from Analytics Help: behavioural modelling:

  1. Consent mode on every page, with Google tags loading in all cases, not only after consent.
  2. At least 1,000 events a day with analytics_storage denied, on at least 7 days, and at least 1,000 daily users with it granted on at least 7 of the previous 28 days.
  3. Other factors, such as the ratio of new to returning users. Meeting the numbers doesn’t guarantee eligibility, and training can take more than 7 days.
  4. Reporting identity set to Blended, not Observed or Device-based (Admin, Data display, Reporting identity; Editor role or above).
  5. No IAB TCF banner for the denial. Google: with a TCF consent solution, “Google Analytics properties aren’t able to model data to fill in the missing information” (consent mode).

Once on, the data quality icon on a report reads “Including estimated user data”, and two things change:

  • Declined visits are counted, not guessed. Google: “Enabling modeling will make unconsented events (analytics_storage=‘denied’) reportable, however these events will remain unaffected by modeling.” Its wording implies they aren’t reportable before that, which is why the table says no events come back under the thresholds.
  • Visitor and session numbers become estimates. GA4 can’t tell that two cookieless hits are one person, so Google says first_visit and session_start run higher for those visitors. Read Users, not first_visit, for audience size.

What does basic mode get?

No visitors back. Google’s comparison table gives behavioural modelling to advanced only, and its text says that if you block tags until consent, “you will not get modeled data in your Google Analytics property to fill in the gaps” (consent mode). The same table marks “Conversion modeling in Google Analytics” for basic, with an asterisk. The footnote says no data is collected when tags are blocked and that conversion modelling “in Ads” uses a general model; Google doesn’t describe the GA4 basic model further. Expect little from it.

Advanced is what lets GA4 estimate declined visitors from your own traffic. Loading Google’s tag before the answer is a privacy-policy decision as well as a tag one; basic vs advanced sets out the trade.

Where will you never see declined visitors?

Even with modelling on, Google lists features it does not apply to (behavioural modelling):

  • audiences and user explorer
  • cohort and user lifetime explorations, retention reports
  • segments with a sequence, predictive metrics
  • data export, BigQuery included (the raw pings are there, the estimates aren’t)

Most exploration templates show consented data only; free-form and segment overlap include estimated users. In path and funnel explorations, modelling is applied to first_visit and session_start, so those counts are lower there than in reports.

Why does BigQuery show more than GA4?

The export holds every raw ping. The reports hold either nothing for declined visitors or an estimate. Google: with consent mode “BigQuery dataset will contain cookieless pings”, “each session will have a different user_pseudo_id”, and “none of the modeled data is available in the BigQuery event export” (Google for Developers: bridge the gap between the UI and BigQuery).

The export also shows your own decline share and whether modelling is possible. privacy_info.analytics_storage reads Yes, No or Unset when consent mode is enabled (BigQuery export schema). For your analyst; swap in your project and property ID:

SELECT
  event_date,
  COUNTIF(privacy_info.analytics_storage = 'No') AS denied_events,
  COUNTIF(privacy_info.analytics_storage IS NULL
          OR privacy_info.analytics_storage = 'Unset') AS unset_events,
  COUNT(DISTINCT IF(privacy_info.analytics_storage = 'Yes', user_pseudo_id, NULL)) AS granted_users,
  COUNT(*) AS all_events
FROM `your-project.analytics_123456789.events_*`
WHERE _TABLE_SUFFIX BETWEEN
  FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 28 DAY))
  AND FORMAT_DATE('%Y%m%d', DATE_SUB(CURRENT_DATE(), INTERVAL 1 DAY))
GROUP BY event_date
ORDER BY event_date;

Count the days denied_events reaches 1,000 (Google wants at least 7) and the days granted_users reaches 1,000 (at least 7 of the 28). granted_users approximates Google’s daily users, close enough to see which side of the line you are. Unset is not always a fault: Google recommends setting the default by region, so traffic outside those regions can carry no state by design. A large Unset share from UK or EEA visitors, or from one page template, is a wiring question. Add geo.country to split it. Other reasons the export and reports differ: GA4 BigQuery export doesn’t match reports.

Consent explains visitors who said no. It doesn’t explain visitors who said yes. Three signs it is wiring, filed as F-201 :

  • Accepted visits are missing too. Accept the banner in a fresh window. No collect request after the accept means the update never reached the tag.
  • A step on launch day. A drop well beyond your decline share, starting the day the banner went live, points to a late default or a page template without the banner code. Checkout and order confirmation are the two I find missing it most often. That is my audit observation, not a Google figure.
  • GA4 flags it. If the property is linked to a Google service that needs consent and a stream has EEA visitors, Admin, Data collection and modification, Consent settings shows whether GA4 receives the advertising and behaviour analytics consent signals. A UK-only or unlinked property may show nothing, which is not an all-clear. Notices can take 48 to 72 hours to update after a fix (Analytics Help: consent settings).

What to do, in order

  1. Compare GA4 purchases with store orders for one closed week, against your decline share, using the rule of thumb above.
  2. Find your row in the table: two questions to whoever set up the banner, or the three-step browser test.
  3. Basic, TCF, or no Consent Mode: declined visitors stay out of GA4 user counts. Whether to change that is the trade in basic vs advanced.
  4. Advanced: set reporting identity to Blended and look for “Including estimated user data”. Nothing after a few weeks: run the query above against both thresholds, and confirm the banner isn’t in TCF mode, before you suspect the set-up.
  5. Accepted visits missing, a step on launch day, or a large Unset share from UK or EEA traffic: that is F-201 , not consent. Run the accept and decline test in how to check Consent Mode v2 is working.

If nobody in-house owns the tags, the Consent Mode service is where I find your row, close F-201 if that is what it is, and show before and after what changed in GA4, including whatever modelling adds once you qualify.

Sources

Read on 16 November 2026:

  • Google Analytics Help, consent mode (analytics storage denied; cookieless pings; comparison table and footnote; basic gets no modelled data to fill gaps; TCF denials can’t be modelled; region-specific defaults)
  • Google Analytics Help, behavioural modelling for consent mode (unconsented events reportable once modelling is enabled; users, sessions and new users; first_visit and session_start; unsupported features; explorations; thresholds and other factors; Blended reporting identity; data quality icon)
  • Google Analytics Help, about modeled key events (key events modelled for unconsented users; core reports; high confidence; Direct on low traffic)
  • Google Analytics Help, DebugView (no debug events without analytics consent)
  • Google Analytics Help, verify and update consent settings (linked properties with EEA visitors; Admin path; two signal sections; 48 to 72 hours)
  • Google Analytics Help, BigQuery export schema (privacy_info fields)
  • Google for Developers, bridge the gap between the Google Analytics UI and BigQuery export (cookieless pings in the export, user_pseudo_id per session, modelled data absent)

← back to the reading room