Cookieless Marketing AI

What "cookieless" means now

For several years marketers were told Google Chrome would remove third-party cookies and that a cookieless future was coming. That plan changed. Google said in 2024 that it would not deprecate third-party cookies in Chrome as planned, dropped the replacement idea of a user choice prompt in 2025, and later said it would retire most of its Privacy Sandbox technologies. Third-party cookies remain available in Chrome.

So why does cookieless marketing still matter? Because the measurement gaps it was meant to prepare you for exist anyway:

• • Consent. In the UK, storing or reading information on a user's device generally needs consent unless an exception applies. Visitors who decline advertising cookies are not tracked in the usual way.
• • Privacy settings and tools. Browser privacy settings, ad blockers and similar tools can stop third-party cookies or tags working for some of your visitors. Browser defaults differ and change, so check each browser vendor's current documentation, and break your own analytics down by browser to see how much of your audience this touches.
• • Platform changes. Ad platforms increasingly rely on data advertisers send them directly rather than on browser cookies.

"Cookieless" in 2026 really means building measurement that does not depend on third-party cookies, because part of your audience is not observable through them anyway, and how large that part is depends on your consent rates and your visitors' browsers.

The UK rules, briefly

The ICO finalised updated guidance on storage and access technologies (cookies, pixels and similar) in April 2026, following changes to the Privacy and Electronic Communications Regulations made by the Data (Use and Access) Act 2025. The default remains that non essential storage and access needs consent. The updated rules add exceptions, including one for statistical purposes, which covers analytics that produce aggregate information used to improve a service. That exception still carries transparency obligations, and it does not cover advertising tracking. Read the ICO guidance itself, or take advice, before relying on an exception.

Where AI fits

AI does three useful jobs in this setting: modelling conversions that cannot be observed, matching conversions to ads using first party data you supply, and estimating channel contribution from aggregate data. None of these works without a sound first party base.

Techniques for measurement without third-party cookies

Technique	What it does	What it needs	Caution
Google consent mode	Adjusts Google tags to each visitor's consent choice and sends signals that allow Google to model conversions from unconsented visitors	A consent banner integrated with Google tags; advanced mode enables better modelling	Modelled conversions appear in reports alongside observed ones; know which is which
Enhanced conversions (Google Ads)	Sends hashed first party data, such as an email address from a form, with conversions so Google can match them to signed in users	Forms that collect the data, consent, and tag or API setup	Only as good as the data captured and the consent obtained
Meta Conversions API	Sends events from your server or CRM to Meta, alongside or instead of the browser pixel	Server side setup or a platform integration; deduplication with the pixel	Duplicate events inflate results if deduplication is wrong
Offline and CRM conversions	Feeds qualified leads and sales back to ad platforms	A CRM with clean stage data and a way to upload	Delays and mismatches between CRM and platform IDs
Marketing mix modelling	Estimates each channel's contribution from aggregate spend and sales over time	Long history of spend and outcome data; statistical expertise	Answers budget questions, not day to day optimisation
Controlled tests	Measures lift by comparing exposed and unexposed groups or regions	Enough volume and a clean holdout	Takes time; small businesses may lack volume

A setup sequence

1. • Audit your consent banner. Does it genuinely block non essential tags until consent, and does it pass the choice to your tags? A banner that sets cookies regardless is a compliance risk and pollutes your data.
2. • Implement Google consent mode, preferably in advanced mode if your legal assessment allows, so modelling has signals to work with.
3. • Turn on enhanced conversions for your main Google Ads conversion actions, using data from your forms or checkout.
4. • Add Meta's Conversions API if you advertise on Meta, and check deduplication with the pixel.
5. • Send qualified outcomes back. Upload CRM stages (qualified lead, sale) to the ad platforms so their AI bidding optimises for real outcomes, not form fills.
6. • Set a budget method. If you spend across several channels at scale, consider marketing mix modelling or regular controlled tests for budget decisions.
7. • Document what is modelled. Keep a note of which reported numbers are observed and which are modelled, and when setups changed.

Illustrative example: reading modelled numbers

A hypothetical retailer sees 100 observed Google Ads conversions in a month. After implementing consent mode in advanced mode, reports show 118 conversions, of which 18 are modelled for visitors who declined consent.

What changed? Not the number of sales in the retailer's till. What changed is how many of them Google can connect to ads, partly by modelling. The retailer should:

• • Compare total sales in its own system before and after, not only the ad platform figure.
• • Expect bidding to shift, because the platform now sees more conversions attributed to ads.
• • Avoid claiming an 18% performance improvement from the change. It is a measurement change.

The figures are round and hypothetical; the size of any modelled uplift depends on consent rates and eligibility, which Google sets out in its own documentation.

What AI cannot fix

AI modelling fills gaps with estimates based on the visitors it can observe. If your tags are broken, your consent integration is wrong or your CRM data is messy, modelling amplifies the error. Fix the base first. And treat any vendor promising to "recover" all lost tracking with caution: some gaps are the result of people choosing not to be tracked, and the law expects that choice to be respected.

How AfM approaches it

In managed work, AfM treats measurement as the first job, because every later decision rests on it. For your own review, the marketing attribution guide covers how to connect spend to revenue, and AI marketing performance metrics covers which numbers to watch.

FAQ

Are third-party cookies going away in Chrome?

Not on current plans. Google said in 2024 it would not deprecate them as planned, dropped a proposed user choice prompt in 2025 and moved to retire most Privacy Sandbox technologies. Check Google's announcements for changes.

Do I still need a cookie banner in the UK?

Generally yes, for non essential cookies and similar technologies such as advertising pixels. The Data (Use and Access) Act 2025 added some exceptions, including for certain analytics, but they carry conditions. Check the ICO guidance.

Is Google's modelled conversion data reliable?

It is an estimate based on patterns from consenting visitors, and Google applies thresholds before modelling. It is useful for bidding and trends; for business reporting, compare with sales in your own systems.

What is the single most useful first step?

Check that your consent banner actually controls your tags and passes consent choices to Google and Meta. Many measurement problems start there, and so do compliance problems.