2026 Google Ads tracking benchmark: what French websites actually send to Google
In short
We read, from the outside, the tracking of 391 randomly drawn French websites with a Google Ads tag. One in three (33%) has Google Ads conversion tags in Google Tag Manager, but none of them sends a value: bidding cannot tell a small sale from a large one. 11.5% still publish Universal Analytics tags, a tool Google shut down in 2023, and a quarter keep at least ten paused tags. Every figure comes with its base and margin of error.
In this study
Key figures
33%
of sites have no Google Ads conversion tag that sends a value
96 of 288 sites · 95% margin: 28 to 39%
11.5%
still publish Universal Analytics tags
45 of 391 sites · 95% margin: 9 to 15%
25%
keep at least ten paused tags in their container
96 of 391 sites · 95% margin: 20 to 29%
57%
at least have enabled enhanced conversions in GTM
163 of 288 sites · 95% margin: 51 to 62%
Light bar: 95% confidence interval (Wilson method). Dark mark: measured value.
One in three without value
Among the 288 sites whose GTM container holds at least one Google Ads conversion tag, 96 have none that sends a value. Each conversion reaches Google Ads with, at best, the fixed value set in the account: the same for a $30 basket and a $30,000 contract.
Smart Bidding optimises what it is given. Without a value, Maximize conversion value and target ROAS have nothing to maximise, and Maximize conversions chases the cheapest volume, often the least profitable.
At the other end, 98 sites (34%) send a value on only some of their conversion tags. That is not necessarily a mistake: a click on a phone number or a newsletter sign-up does not always have an obvious value. What matters is that the conversion driving bidding carries one.
To give your conversions a value, lead generation included, see our guide to profit-based bidding.
Universal Analytics, still there
Google stopped processing Universal Analytics data on July 1, 2023, and on July 1, 2024 for Analytics 360. Yet 45 of 391 sites (11.5%) still publish at least one Universal Analytics tag wired to a trigger, and 16 of them send a Universal Analytics request as soon as the home page loads.
These tags measure nothing anymore. Above all, they show that the tracking plan has not been reviewed since the move to GA4: that is often where conversions counted twice, or tied to the wrong event, are hiding.
The state of GTM containers
Google Tag Manager remains the norm: 88% of retained sites load it, and 74% declare their Google Ads conversion tags there. The others use a tag coded into the page, an import from GA4 or server-side tagging, all invisible from outside.
| Finding | Sites | Rate | 95% margin |
|---|---|---|---|
| GTM loaded on the home page | 344 of 391 | 88.0% | 84 to 91% |
| Google Ads conversion tag in GTM | 288 of 391 | 73.7% | 69 to 78% |
| Enhanced conversions enabled in GTM (at least) | 163 of 288 | 56.6% | 51 to 62% |
| At least ten paused tags | 96 of 391 | 24.6% | 20 to 29% |
| More than one GTM container loaded | 23 of 391 | 5.9% | 4 to 9% |
The enhanced conversions figure is a minimum: they can also be enabled in the Google tag or sent through the API, which reading the container cannot see. Paused tags are housekeeping rather than errors, but one site in four keeps at least ten of them (the median is two): old versions that make every change harder. Several containers on one page are not a fault in themselves (agency and brand, several countries), but they are the first cause of conversions counted twice.
Method
Population. The .fr domains of the Tranco list (an academic popularity ranking, version Y897G): 10,582 domains. 2,000 were drawn at random (seed 20261115), our own clients' sites excluded.
Reading. One load of the home page per site, in an automated browser, with no click and no form, then a read of the published GTM container, which is public by design. 1,353 sites could be read (68%); the others were offline, protected against bots or too slow.
Retained sites. A site enters the study when a Google Ads ID (AW-) or a Google Ads conversion tag is visible: 391 sites. Excluded: sites that only sent a generic advertising request (often a publisher's ad network), sites redirecting to another domain, and duplicates.
| Step | Sites |
|---|---|
| Drawn at random | 2,000 |
| Read | 1,353 |
| With a Google Ads tag, retained | 391 |
| Of which a Google Ads conversion tag in GTM | 288 |
Reading rule. From outside, only the presence of something is certain. A figure is published as a fact only when it rests on positive evidence (a published Universal Analytics tag, no value field on any tag read); settings that can live elsewhere are given as a minimum. Each rate comes with its base and its 95% confidence interval (Wilson method). The value reading was checked by hand on 20 randomly drawn sites, with no discrepancy; every site flagged for Universal Analytics or multiple containers was reviewed.
Data. No site is named. The per-site file stays offline and will be neither published nor used for prospecting.
What we did not publish
- The gclid lost on arrival. The first pass showed 16% of sites losing the Google Ads click ID. Reviewed one by one, almost all lost it only when moving from the address without "www" to the address with "www": an ad usually points straight to the latter. Of 77 cases retested, 76 kept the gclid. The figure was withdrawn.
- Consent. Banners and Consent Mode behave differently depending on the visitor's country; this edition does not measure them.
- Whether values are right. The study sees whether a tag sends a value, not whether that value is correct.
- Sites without a Google Ads tag, and French brands on
.com, are outside the population by design. The Tranco list also favours high-traffic sites.
FAQ
How many websites were analysed?
2,000 .fr domains were drawn at random, 1,353 could be read, and 391 had a Google Ads tag. Rates are based on these 391 sites, or on the 288 that declare Google Ads conversion tags in Google Tag Manager, depending on the indicator.
Is a conversion tag without a value always a mistake?
No. A micro-conversion, such as a click on a phone number, can do without one. The problem starts when no tag on the site sends a value: value-based bidding then has nothing to optimise, and every conversion looks the same to Google Ads.
Why are Universal Analytics tags still published?
Because the move to GA4 was often done by adding tags without removing the old ones. They have measured nothing since 2023, but they point to a tracking plan nobody has reviewed.
Is my website in the sample?
No site is named and per-site results are not published. To see what your own site sends to Google Ads, the tracking self-check runs the same reading in two minutes.
Can I quote these figures?
Yes, citing "TAGBRUT Google Ads tracking benchmark, November 2026" with a link to this page. Each figure must keep its base: 391 French websites with a Google Ads tag.
What about your site?
The tracking self-check runs the same reading as this benchmark on your site, in two minutes, then asks four questions about what cannot be seen from outside.
To go further, the audit is free: the five most costly findings, quantified, in writing within 48 business hours.
Study by the TAGBRUT team.
Sources (4)
- Tranco (tranco-list.eu), list Y897G: a domain popularity ranking built for research (Le Pochat et al., NDSS 2019).
- Google Analytics Help, "Google Analytics 4 has replaced Universal Analytics": data processing stopped on July 1, 2023, and on July 1, 2024 for Analytics 360.
- Google Ads Help, "Conversion Values Best Practices": conversion values and value-based bidding.
- Google Ads Help, "About enhanced conversions": set up through Google Tag Manager, the Google tag or the API.