Tag: Identity resolution

All blog posts with this tag.

Safari blocks identity-resolution services

Safari blocks identity-resolution services

On 19 October 2026, WebKit announced protection against certain identity-resolution infrastructure used to associate the same person with activity across different websites. The post describes third-party services that can receive an email address during a sign-in or subscription and map it to tracking identifiers. This is a useful development for teams evaluating "cookieless" analytics. Removing third-party cookies does not, by itself, eliminate techniques for identifying or connecting people. Other mechanisms can start with information someone provides for a legitimate purpose, such as receiving a newsletter, before a third party uses it for a different purpose. The opposite overstatement is equally misleading: the announcement does not say Safari blocks all analytics or all third-party services. WebKit describes a particular class of infrastructure. Understanding the scope matters when reviewing a website's data flows. What WebKit actually announced in October In "Blocking Identity Resolution Infrastructure", WebKit explains how an identity-resolution provider may obtain an email address on sign-in or registration, create or look up a linked identifier, and return that identifier to a page. The provider may then connect activity across websites. WebKit says Safari and WebKit in the version 27 Apple operating-system releases block network access to identified identity-resolution infrastructure. Its description concerns domain-level network blocking, rather than simply removing a particular request parameter. This is a browser engineering announcement, not a change to European law. It does not establish that any specific analytics tool has stopped functioning. Product impact requires inspection of the domains and mechanisms actually used. How identity resolution works Consider a visitor using the same work email to download a report from one website and sign in to another service. Each website has a separate legitimate reason to receive that email. A third-party provider embedded in both may try to identify the same person and assemble their interactions. This does not require a shared third-party cookie. Linking can occur through email-derived identifiers, opaque keys or associations computed on the provider's servers. Different uses deserve separate treatment:Situation Primary purpose Review questionAccount sign-in Access a requested service What is needed for authentication?Newsletter or contact form Process a specific request Which recipients receive fields?Aggregated audience measurement Understand website use Are identifiers and purposes documented?Cross-site identity resolution Link personal activity across contexts What authorises the linking and how is it explained?These activities can coexist within one technical stack. The fact that a website needs an address for a contact form does not authorise every secondary use of that address. Why "cookieless" is an incomplete description Cookieless may describe the absence of cookies for one operation. It does not reveal whether persistent server-side identifiers are created, whether email values are transformed into identifiers, whether behaviour is joined across properties or whether third parties receive events. Separate three questions: storage mechanism, ability to identify or link someone, and purpose of the processing. Treating them as synonyms creates both engineering and legal confusion. One tool may collect few values but still transmit events. Another may use identifiers without a cookie. A third may offer multiple collection modes. Marketing vocabulary cannot replace testing the configuration in use. The CNIL describes narrow, conditional possibilities for consent exemptions in some audience-measurement settings. That is not a universal exemption for every cookieless system, especially where marketing or cross-site profiling purposes are introduced. A practical review for marketing and engineering A small team can run a useful audit without trying to identify real visitors. The goal is to document flows and responsibilities. Map third-party providers Inventory form, newsletter, booking, analytics, testing and personalisation scripts. Record receiving domains, intended purposes, the accountable internal owner and fields likely to be transmitted. Plugins can load additional resources, so check network traffic rather than just plugin names. Use test pages and fictional contact details. Inspect forms and sign-in Use a synthetic email address in a test form. Watch requests during typing, submission and confirmation. The question is not only "is there a cookie?" It is "which data leave this origin, and for what service?" Scripts may also read query parameters. Our URL filtering guide explains why unnecessary URL fields can leak identifiers. Do not use real credentials or production tokens in this test. Compare browsers without overstating results Repeat the journey in Safari and another browser, recording versions and privacy settings. Differences can arise from browser protection, content blockers, user choices, configuration or ordinary bugs. One missing request does not establish the precise reason it was blocked. Record requests and responses, and ask the vendor for an explanation before changing controls. Separate processing purposes Newsletter subscription and audience measurement answer different questions. Cross-site identity linking introduces another form of processing. Vendor documentation should make those uses distinguishable. If a tool that claims to count visits receives a contact email, that inconsistency deserves a closer examination. A legitimate operational need should be explainable. Does browser blocking reduce actual traffic? Not necessarily. A protection may stop requests to a third party without changing the number of people who visited the website. Its impact on a dashboard depends on where collection occurs and how the metrics are built. Keep three observation layers distinct: HTTP requests seen by the service, browser-side analytics events and reported aggregates. They can differ because of blocking, robots and sessions without JavaScript. An apparent reporting drop is not proof of a commercial traffic drop. Verify definitions and compare relevant independent signals before drawing that conclusion. Multi-site governance does not require an identity graph A company with marketing, blog and documentation properties often needs common reporting: acquisition sources, useful content, trends and shared ownership. Those needs do not automatically require recognising the same individual across all properties. Property-level and period-level measurement can remain actionable without building cross-site identities. Our multi-site analytics guide shows how to structure shared reporting without making personal tracking a prerequisite. What the announcement does not change WebKit's protection does not replace vendor contracts, processing documentation, user information or configuration audits. Nor does it imply that every third-party service is inappropriate. The technical lesson is more specific: information collected for a clear first-party purpose can become intrusive when repurposed to link otherwise separate contexts. Pomelo's "cookieless by default" positioning describes a design direction. It is not a general legal exemption or a universal compatibility promise. Teams should verify the implementation and current configuration they actually use. Questions to put to an identity or analytics provider A supplier review should focus on mechanisms rather than generic assurances. Ask whether the service receives raw email addresses or transformations of them, whether it issues identifiers that survive between sessions, and whether different customers' website data can be linked. Distinguish an identifier created by the first-party service for login from one designed to recognise someone across separate websites. Ask how the vendor documents data recipients and technical sub-processors. Where an identifier is transformed, check who holds any mapping or additional information necessary to interpret it. Hashing a known address is not by itself proof of anonymisation, because the same input may still be recognisable through comparison or additional data. Evaluate purpose boundaries separately from technical controls. A company may legitimately need an email to send a requested document. That does not settle whether the address can be passed to an identity-resolution provider for advertising or cross-site profiling. Finally, test the operational consequences of blocking. If a browser prevents one third-party call, does the form still succeed? Does the website continue to provide the requested service? A well-designed service should not rely unnecessarily on cross-site tracking to perform a basic user-requested action. The outcome should be a short data-flow map, not an unsupported certificate that the site is "tracking-free". Conclusion: audit flows, not slogans The October Safari update provides a straightforward question for any stack: can the team explain where an email entered into a form goes, and who can connect it to other activity? Web teams do not have to identify every person across multiple sites to understand content and acquisition. They need to know what is transmitted, limit purposes and justify choices. That principle outlasts any one browser release. FAQ Does Safari block every analytics tool? No. WebKit describes blocking of identity-resolution infrastructure, not all audience measurement. Is a hashed email automatically anonymous? No. A transformed value can still support linking or reidentification depending on the method and available data. The actual mechanism matters. Does cookieless mean consent is never needed? No. Applicable requirements depend on purpose, technology and legal context. Cookie absence alone is not enough. Is testing Safari alongside another browser useful? Yes, provided versions and settings are recorded. Differences should be investigated, not bypassed without justification. SourcesWebKit, Blocking Identity Resolution Infrastructure, 19 October 2026 WebKit, Tracking Prevention Policy CNIL, audience-measurement tools and consent conditions CNIL, what the law says about trackers CNIL, personal data, pseudonymisation and hashing (French)

Just the right measure, for all your websites.

Pomelo is opening gradually. Join the first teams measuring right and help shape what comes next.