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

READING ROOM › FIELD NOTE › NOTE-011

NOTE-011 · FIELD NOTE FILED 08 NOV 2026 · REVISED 08 NOV 2026 F-201

What a Consent Mode v2 audit covers and what it hands back.

ABSTRACT — A Consent Mode v2 audit tests whether visitors who reject your cookie banner are treated as rejected, whether Meta and other tags wait for the answer, and whether a conversion drop since the banner went live is refusals or wiring. What it checks, what you should get back, and a reconstructed test record.

Baris Asa, Synctag, tracking specialist, London · Last checked: November 2026

A Consent Mode v2 audit answers three questions. Are visitors who reject your cookie banner really treated as rejected? Do Meta and your other ad tags wait for the answer? And is a conversion drop since the banner went live caused by people saying no, or by the wiring? It should hand back screenshots of what each tag was told after accept, reject and a partial choice, coded findings with evidence and a fix, and a check of the drop against your banner’s refusal rate.

Synctag is my London practice. As a Consent Mode consultant I audit and repair Consent Mode v2 for UK online shops and the agencies behind them, and every audit is tested on reject as well as accept. I have no partnership or commission with any banner vendor. The audit is read-only and takes a day or two for one site and one banner.

Do you need one?

Two quick signs, both from your own accounts:

  • Conversions fell from the day the banner went live, by more than the share of visitors the banner dashboard says rejected.
  • Google Ads doesn’t say “modeling is active”. Open Goals, then Conversions, then Summary, pick a website conversion action and read its Diagnostics tab (Google Ads Help).

Neither proves a fault. A wiring fault can also show no drop at all, as in the exhibit below. Either sign is a reason to test, not a diagnosis. The fault, when it is there, is on record as F-201 .

To run the browser checks yourself, see How to check Consent Mode v2 is working. For why conversions drop after a banner goes live, see Conversions dropped after implementing Consent Mode v2. This note covers the audit and what you get back.

Five parts, in the order a tag meets them. Consent Mode v2 sends Google four switches: ad_storage, analytics_storage, ad_user_data and ad_personalization (Google’s guide). The evidence for most parts is the Consent tab in Google’s free Tag Assistant tool: the panel that shows what each switch was set to, and when.

PartThe business questionEvidence
1. DefaultsIs every switch set before any tag fires?Consent tab, before any click
2. The updateDoes reject really turn things off?Consent tab after each choice
3. RegionsDo UK visitors get the setting you chose?Consent tab from two locations
4. Other tagsDo Meta and TikTok wait for the answer?Tag settings, network hits
5. Google’s viewCan Google model what you lose?Ads Diagnostics, GA4 settings

1. Defaults. If a tag fires before the switches are set, it runs on no instruction at all. Google says the default has to run before any command that sends data: “if your consent code is called out of order, consent defaults won’t work” (Google’s guide).

In Google Tag Manager, the tag that sets consent belongs on the Consent Initialization trigger, which fires before every other trigger (consent settings). The audit records each default and when it landed.

2. The update. This is where a visitor’s “no” either reaches the tags or doesn’t. The banner’s categories have to map onto the four switches: “marketing” commonly to the three ad switches, “statistics” to analytics_storage. The mapping is the banner’s set-up, so the audit writes down what it actually is.

The partial choice is the one to test: statistics accepted, marketing refused. A switch missing from the update hides there, because accept-all and reject-all can both look right.

3. Regions. A site can set one default for the UK and another for everywhere else. Google’s regions use ISO 3166-2 codes, and the more specific region wins (Google’s guide). The UK’s code is GB; UK is not an ISO code, and Google’s guide gives it no defined behaviour.

Each region row needs all four switches. Google’s guide doesn’t say what happens to a switch a region row leaves out, so the audit reads the Consent tab rather than assume. If the banner loads late, wait_for_update (in milliseconds) sets how long Google’s tags wait for it; the audit checks it suits your banner.

4. Other tags. Google’s tags adjust to Consent Mode themselves (GTM calls these built-in consent checks). Meta’s pixel has no built-in check in GTM, and other vendors’ templates vary, so each is checked. A tag without one needs the setting “Require additional consent for tag to fire”, or a trigger from the banner. GTM’s Consent Overview lists every tag still marked “Consent Not Configured” (consent settings).

Tags pasted into the site’s code outside GTM are beyond any container setting. The audit lists them.

5. Google’s view. In Google Ads, each website conversion action shows “Consent mode is implemented” or “Consent mode is implemented and modeling is active” (Google Ads Help). In GA4, the audit reads whether the property meets the modelling thresholds and whether reports use the Blended reporting identity, the only setting that shows modelled data (GA4 Help). If a server container is in the path, consent has to reach it too: Consent Mode v2 with server-side GTM.

What should the audit hand back?

Six things. If one is missing, ask for it.

  1. A consent test record. The Consent tab for default, accept, reject and a partial choice, in each region tested, as screenshots.
  2. A mapping table. Each banner category, the switches it sets, and the tags it governs. Your developer makes the fixes from this page.
  3. Findings. One per fault, coded (F-201 for consent wiring), with evidence, the conversions it affects, who fixes it, the fix and a re-test. The fields are set out in what a tracking audit report should hand back.
  4. The drop check. Store orders against Google Ads and GA4 conversions either side of the banner date, set against the refusal share in the banner’s dashboard (method below).
  5. The modelling status. Whether you run basic or advanced mode, and which of Google’s thresholds the account and property meet today.
  6. What wasn’t tested. A market with its own banner, an app, a server container nobody gave access to. Unchecked is on record as unchecked, not as passed.

Two optional settings get a written decision rather than a default: url_passthrough, which carries ad-click information across pages in the URL, and ads_data_redaction, which removes ad-click identifiers from requests when ad_storage is denied (Google’s guide). Whether each is on is your call; the audit says what each one changes.

The finding in this exhibit: a GB default that set only ad_storage and analytics_storage left ad_user_data and ad_personalization granted after reject.

Exhibit: one site, GB region, fresh Incognito visit. Faulty cells in bold.

SwitchDefault (GB)After rejectAfter accept
ad_storagedenieddeniedgranted
analytics_storagedenieddeniedgranted
ad_user_datagrantedgrantedgranted
ad_personalizationgrantedgrantedgranted

Reconstructed · client data anonymised

The GB row set only the two original switches. On the Consent tab, UK visitors got the two v2 switches from this site’s global row, which set them to granted, and the banner’s reject update didn’t touch them. Visitors who refused were still sent to Google as agreeing to ad data use. Filed as F-201 .

Fix: add both switches to the GB default and to the reject update. Re-test: the same three columns, all four denied after reject.

On accept, this site looked correct. Its conversions didn’t drop either: the fault failed open. Only the reject test found it.

It compares store orders with Google Ads and GA4 conversions for the weeks either side of the banner date, and sets the drop against the share of visitors the banner dashboard says rejected. A drop that starts on the banner date and is clearly bigger than the refusal share is a reason to check the wiring first. It is not proof: seasonality, a tag change in the same week, or basic mode can produce the same shape. The Consent tab record decides.

How much fixing the wiring brings back as modelled data depends on the mode. In basic mode Google Ads uses a general model; advanced mode gives an advertiser-specific one (Google’s consent mode overview). Smaller accounts may never get modelling, and the Google Ads Diagnostics tab tells you where you stand today.

WhereWhat Google needs before it models
Google Ads700 ad clicks over 7 days, per domain and country
GA4Advanced mode on every page; 1,000 events a day with analytics_storage denied for 7 days; 1,000 daily users with it granted on 7 of the last 28 days

Sources: Google Ads Help, GA4 Help. GA4’s model needs Google tags that load whether or not the visitor consents, which is advanced mode, and Google says meeting the numbers doesn’t guarantee it. The full trade-off is in basic vs advanced. The audit says which side of each line you are on; it doesn’t promise a figure.

  • A legal review. Which cookies need consent under UK PECR is a call for your data protection lead or lawyer. The audit records that decision and tests that the tags follow it.
  • A banner design review. Wording and button layout belong to the banner vendor and your lawyer.
  • A change to your live tags. It is read-only. Repairs come after, as named GTM versions published by whoever owns the container.
  • A reason to switch banner. A new banner sends the same switches through the same container; if the wiring is wrong, it stays wrong. How the main banners compare: four banners compared.

What should you ask before hiring someone for it?

Google’s EU user consent policy covers visitors in the EEA, the UK and Switzerland (Google’s policy). Three questions specific to a Consent Mode audit:

  1. Do you test the GB row for all four switches, from a UK location?
  2. Do you test a partial choice, not only accept and reject?
  3. Do you check Meta and the other non-Google tags as well as Google’s?

The wider question of who handles privacy-first tracking and consent in the UK is in Privacy-first tracking in the UK.

Common questions

No. It tests whether the tags do what the banner says. Which cookies need consent under UK PECR is a decision for your data protection lead or lawyer; the audit records that decision and tests that the tags follow it.

The audit changes nothing by itself; it is read-only. In advanced mode, fixing what it finds lets Google model conversions from visitors who decline, once the account meets Google’s thresholds. In basic mode, tags don’t load before consent: Google Ads falls back to a general model and GA4 behavioural modelling doesn’t apply. The audit tells you which case you are in.

A day or two for one site and one banner; more banners or regions add time. Read access to Google Tag Manager, Google Ads and GA4, and a login to the banner’s dashboard. Nothing is published during the audit.

The wider picture, with every Consent Mode question in one place, is in Google Consent Mode v2 explained.

Sources

Read on 8 November 2026:

  • Google for Developers, set up consent mode on websites (v2 switches; default before data is sent; ISO 3166-2 regions, more specific wins; wait_for_update in milliseconds; url_passthrough; ads_data_redaction)
  • Tag Manager Help, consent settings (built-in and additional consent checks; Consent Overview; Consent Initialization trigger)
  • Google Ads Help, set up and manage consent (Diagnostics statuses; 700 ad clicks over 7 days per domain and country)
  • Google Analytics Help, behavioural modelling for consent mode (tags load regardless of consent; consent mode on all pages; 1,000 denied events a day for 7 days; 1,000 granted daily users on 7 of 28 days; no guarantee; Blended reporting identity)
  • Google for Developers, consent mode overview (basic: general model; advanced: advertiser-specific model)
  • Google, EU user consent policy (EEA, UK and Switzerland)

— B.A., filed from London

CONNECTED RECORDS F-201

REVISION HISTORY — revised 08 NOV 2026.

← back to the reading room