Advertising without third-party cookies: contextual, first-party data and alternative IDs

For years the industry treated the "end of third-party cookies" as a single date. It never was. For an exchange operator the useful question is narrower: which signals arrive in the requests you handle, which ones your demand partners can use, and how you make sure nothing valuable gets lost on the way.

Where things actually stand

  • Safari blocks third-party cookies by default, for all users since 2020, on top of Apple's Intelligent Tracking Prevention.
  • Firefox isolates cookies per site by default with Total Cookie Protection, so a third-party cookie cannot follow a user across sites.
  • Chrome changed course: Google dropped its plan to phase out third-party cookies, and in 2025 retired most of its Privacy Sandbox advertising APIs. Third-party cookies stay on by default, and users can block them in settings.
  • Apps never used cookies. Device advertising IDs play that role, and they are limited too, most visibly on iOS where access needs the user's permission.

So a large part of the traffic passing through any exchange already carries no cross-site cookie ID, and consent rules such as the GDPR apply whatever the browser does. Exchanges that depend only on cookie matching are selling to a shrinking slice of their own supply.

Contextual signals

Contextual targeting places ads based on the content being viewed, not on who is viewing it. It needs no personal data, and modern classification can read pages and apps far more precisely than keywords did.

To make context machine-readable, the IAB Tech Lab maintains a Content Taxonomy, a standard list of content categories. OpenRTB 2.6 added a cattax field so a request can say which taxonomy and version its category codes use.

For an operator: contextual buyers can only bid on what reaches them. Pass on site and app data completely: domain or bundle, page or store URL, categories with their cattax, placement and format information. If a supply partner sends categories without saying which taxonomy they use, ask, so buyers do not read the codes against the wrong list.

First-party data

First-party data is what a publisher learns directly from its own users, with consent: logins, subscriptions, on-site behaviour. It does not depend on third-party cookies, and it belongs to the publisher.

Publishers usually monetise it through deals: segments are matched on the publisher's side and the deal ID travels in the request, so the raw data does not have to leave their hands.

For an operator: make sure deal information passes through your exchange untouched and reaches the right buyers. A deal that breaks somewhere in the chain looks to everybody like a deal that does not work.

Alternative identifiers

Alternative IDs try to recognise the same user across sites without third-party cookies. Broadly:

  • deterministic IDs, usually derived from a hashed and encrypted email address the user provided when signing in, with Unified ID 2.0 as a well-known example;
  • probabilistic or shared IDs, which combine signals available to a publisher to estimate whether two visits come from the same user, with ID5 as one provider in this space.

In OpenRTB these travel in the eids array on the user object: user.eids in 2.6, user.ext.eids in older integrations. Each entry names its source and lists the IDs from it.

For an operator: pass eids on to the demand partners who use them, in the place their integration expects, and respect consent signals while doing so. Do not invent or overwrite IDs, and do not send identifiers to partners who are not supposed to receive them.

Consent

Signals are only usable if they were collected lawfully. Consent and privacy information arrives in the request from the publisher's side. Your exchange's job is to keep it intact and to make sure partners act on it. Check this during partner onboarding, not after a complaint.

What to do now

  • Find out how much of your traffic already arrives without any cookie-based user ID.
  • Audit what your supply partners actually send: categories, taxonomies, deal IDs, eids, consent fields.
  • Make sure every one of those reaches the demand partners who can use it.
  • Ask demand partners which signals they bid on, and prioritise those.

An exchange does not create audience data. What it can do is carry every useful, lawful signal from seller to buyer without losing it. In a market with fewer cookies, that is a real reason for partners to choose you.

Want to talk about your own setup? Get in touch with the Xendiz team.