Which Google Consent Mode Is Your Site Actually Running?

Most practice websites cannot answer which Google Consent Mode they are running, and the reason is that nobody chose it. It is not a setting in a dashboard. It is a property of how the tags were written into the page, decided years ago by whoever installed them, and frequently different for different tags on the same site.

Working out which one you have takes about two minutes and changes what you can honestly say about your own site. It also explains something that otherwise looks inexplicable, which is why two practices running what appears to be the same consent banner, from the same vendor, configured by the same agency, can behave completely differently when you actually watch the network.

Two modes, one real difference

Google Consent Mode v2 has two implementation patterns, and the difference is not what the banner looks like. It is whether Google's tag is on the page before the visitor decides anything.

Advanced. The tag loads on page load. Before consent it sets all four v2 parameters to denied and sends cookieless pings: no identifiers, no cookies, just a signal that a page was viewed. When the visitor consents, a second call flips the relevant parameters to granted and normal measurement begins. Those pre-consent pings are what Google uses to model the conversions it could not observe directly.

Basic. The tag never loads until the visitor consents. Nothing is sent beforehand, cookieless or otherwise. When consent arrives, the tag loads and fires normally. There are no pre-consent pings, so there is nothing to model from.

Which Google Consent Mode you have is an install property

This is the part that surprises people. You do not pick Basic or Advanced anywhere. You pick it by how the tag is written into the page.

A tag written the conventional way is fetched by the browser immediately, which means you are running Advanced at best:

<script async src="https://www.googletagmanager.com/gtag/js?id=G-YOUR-ID"></script>

A tag written as an inert placeholder is not fetched at all until a consent manager rewrites it, which is Basic:

<script type="text/plain" data-consential-category="analytics" data-consential-src="https://www.googletagmanager.com/gtag/js?id=G-YOUR-ID"></script>

The second pattern works because a browser will never execute a script whose type is not a recognised JavaScript MIME type. The consent manager finds those placeholders and converts them into real script tags once the matching category is granted. It is the standard client-side blocking technique, and its useful property is that blocking becomes independent of load order.

Why most sites run both Google Consent Mode patterns at once

Because the mode is a property of each individual tag rather than of the site as a whole, a single practice website routinely ends up running both patterns simultaneously without anyone having decided to.

A GA4 tag converted to the placeholder convention is blocked. A Meta pixel injected directly by a plugin, that nobody converted, is not blocked at all, and no consent manager can intercept what it never had a chance to rewrite. Mixed installs are the norm rather than the exception, and usually nobody involved knows it.

This is also the mechanism behind a claim worth internalising: a consent banner is not evidence of blocking. A banner can be present, correctly worded and fully functional while the tags behind it fire on page load regardless of what the visitor clicks. The general version of that argument is in consent mode is not tag blocking, and where the unconverted tags tend to hide is covered in where tracking tags actually load.

The reason mixed installs persist is that converting a tag requires somebody to know it exists. A consent manager can only rewrite the placeholders it was given, and a plugin that writes its pixel directly into the page at render time never presents one. So the tags most likely to be governed are the ones a technical person deliberately handed over, and the tags least likely to be governed are the ones that arrived through a settings screen somebody filled in once.

Why we default to Basic

Consential sets all four parameters to denied in every US jurisdiction, including states where a granted default would be permitted, and holds Google's tag until the visitor acts.

The reason is context. A cookieless ping carries no identifiers, which on a retail site makes it genuinely low risk. On a practice site the ping still carries the page address, and the page address may name a procedure. The identifier is absent and the health context is not, which is a materially different situation from the one Advanced was designed around. That reasoning is developed in Consent Mode v2 for healthcare advertisers.

The cost is real and worth stating plainly. Basic gives up modelled conversions, so reported conversion counts in Google Ads will be lower than under Advanced. That is a measurement loss, not a performance loss: the same patients still book, and fewer of them are attributed. For practices where most conversions are phone calls or form submissions handled server side, the practical gap is smaller than the headline suggests, because the modelling was estimating browser-side conversions that were not the whole picture anyway.

There is a second-order effect worth anticipating, because it causes alarm when nobody warned about it. Smart bidding strategies optimise against reported conversions, so a switch from Advanced to Basic reduces the signal those strategies receive and campaigns typically need a learning period to resettle. The right way to make the change is deliberately, at a point in the month where a fortnight of noisier performance is tolerable, with the date recorded so that a later dip is attributed correctly rather than chased as a campaign problem.

The alternative framing some practices prefer is that the modelled numbers were never patients in the first place. A modelled conversion is an estimate of an event nobody observed, which is a legitimate input to a bidding algorithm and a poor input to a conversation about how many consultations the practice booked. Losing them removes an estimate rather than a patient, and practices that reconcile ad platform figures against their actual calendar tend to find the reconciliation easier afterwards, not harder.

How to check your own site in two minutes

Two checks, and neither needs access to anyone's implementation.

First, view source and search for googletagmanager.com. If you find a conventional script tag with a live src, that tag is not blocked. If you find type="text/plain" with the source in a data attribute, it is. Do this for every third-party script you find, not just the first, because the answer varies per tag.

Then check behaviour rather than markup. Open a private window with the network panel open, load a treatment page, and touch nothing. Requests appearing before you interact with the banner tell you what is actually happening, which is the authoritative answer when the markup is ambiguous. Google's implementation guide describes what a correctly wired default call should look like if you want to compare.

Behaviour beats markup whenever the two disagree, and they disagree more often than you would expect. A tag can be correctly written as a placeholder in the theme and then injected a second time by a plugin in its original form, so the source shows the careful version and the network shows the careless one. A caching or optimisation layer can combine scripts and defeat a rule that identified them by filename. In both cases the markup is telling you what somebody intended and the network is telling you what happened.

One last thing to check while you have the panel open: whether the requests stop after you decline. Some implementations block correctly before a decision and then release everything on any interaction, including a dismissal, which converts a considered no into an accidental yes.

Whichever mode you conclude you are running, the useful output of this exercise is the list of tags that turned out not to be governed by anything. That list is almost always longer than expected, and it is what consent gating is for.