AI search attribution and privacy:
what to collect, and how
AI search attribution can run on first party data you already collect. Server logs hold IP addresses, which count as personal data under GDPR. A source cookie may need consent. CRM fields and survey answers need a clear notice. This lesson is general guidance, not legal advice.
Four attribution signals each touch personal data in a different way
Diagram of privacy handling by signal. Not legal advice.
Diagram. General guidance, not legal advice. Check with your counsel and your region's rules.
How to read it. Read across a row. The cheapest signal to handle is the survey answer. The one to treat with the most care is the server log, because it holds IP addresses.
In ThriveStack citedby Attribution runs on first party data you collect, with no bought data match. See it for your site →
Attribution runs on first party data, so start from what you already collect.
You do not need bought data. Referrers, UTM values, logs, CRM fields and survey answers are enough.
- Referrer and UTM values on the visit
- Server log lines for bots
- A source field on the CRM contact
- An optional survey answer
Server logs hold IP addresses, which count as personal data under GDPR.
Treat logs as personal data. Keep them only as long as you need them.
- 1
Shorten or hash IP addresses
Remove what you do not need.
- 2
Set a retention period for logs
Delete old logs on a schedule.
- 3
Document how you handle log data
Write the purpose in your privacy notice.
A source cookie needs consent in many regions.
Store the source fields only, and ask for consent where your region requires it.
- Referrer
- utm_source and utm_medium
- Landing page
- Anything you cannot explain
- Bought identity data
- More than you need
Surveys and CRM fields need a clear notice.
Tell people why you ask, and honour deletion requests.
Check your own region's rules with counsel. This guide is general guidance.
ThriveStack citedby runs your prompt panel across 10+ AI engines, reads server-log crawler hits, matches human visits to signups and reports AI-sourced and AI-influenced pipeline.
AI attribution privacy, answered.
Is AI search attribution GDPR compliant?
It can be, if you collect only what you need, say why and honour rights. Check with counsel.
Are IP addresses personal data?
Under GDPR, yes.
Do I need consent for a source cookie?
Often, depending on region and what the cookie stores.
All 10 questions this lesson answers
Is AI search attribution GDPR compliant?
It can be, but it depends on what you collect. A referrer and a page view are low risk. IP addresses can count as personal data under EU case law (Breyer, 2016) when you have legal means to identify the visitor. Keep identifiers short lived and ask a lawyer about your setup. This is not legal advice.
How to track AI referrals without cookies
Read the referrer and UTM values on each page view, either in the browser or on the server, and store only the source and landing page. Cookie free tools such as Plausible do this. The tradeoff is that you cannot link a person's later visits.
Consent mode and AI referral tracking
When analytics_storage is denied, GA4 sets no analytics cookies. In Advanced consent mode it still sends cookieless pings that Google can use for modeling. Basic mode blocks the tag fully. Modeling needs data thresholds and is not guaranteed.
Privacy safe AI attribution
Collect the source, the landing page and the time, not the identity. Aggregate before you report. Trim or drop IP addresses, set short retention, and honor consent choices. Use the CRM only for people who gave you their details through a form.
Do AI crawlers respect cookie consent?
Crawlers never see your banner and do not set cookies. They fetch pages as bots and do not run your analytics script. So consent applies to people, while bot hits show up only in server logs. Robots.txt, not a banner, is how you set rules for bots.
Server side tracking for AI traffic
Server logs and server side tagging catch visits that scripts miss, such as AI fetches that never run JavaScript. Matomo, for example, uses server side tracking to see chatbot requests. If you keep IP addresses or IDs, privacy rules still apply. ThriveStack citedby reads crawler data from Cloudflare, Vercel and WordPress.
First party data for AI search attribution
Capture the AI source on your own site, then store it with the contact when they fill in a form. That data is first party and consent is clear. ThriveStack Revenue Intelligence joins website visitors with CRM and billing records you already own.
Can you track ChatGPT users without personal data?
Yes, at the group level. A visit from chatgpt.com, or one tagged utm_source=chatgpt.com, tells you the platform and the page. Counting those visits needs no name, email or ID. You only need personal data to follow one person across visits.
AI attribution and CCPA
CCPA covers a business that earns over $25 million (adjusted to $26.625 million from 2025), handles data on 100,000 or more consumers or households, or earns half its revenue from selling or sharing data. Plain analytics is not the issue. Sharing identifiers with ad networks is, and honoring Global Privacy Control matters.
Cookieless attribution for AI search
Rely on referrer and UTM values per visit, server logs for bot activity, and a self-reported field in your forms. You lose the link between visits, so expect a rougher picture. Consent mode modeling can fill some gaps in GA4.
Sources, checked November 2026: Google: Consent mode overview · Google: Set up and manage consent · CJEU Breyer summary (SCL) · CCPA applicability (Clym) · Matomo AI Reports Guide · Plausible: AI traffic.