Session replay is one of the few analytics tools that records what a person types, so a session replay GDPR review is less about whether the tool is allowed and more about how you configure it. The regulation doesn’t ban recordings; it asks for a lawful basis, data minimisation, a retention limit and a way to erase.
Most teams do this backwards: install the recorder, ship it, then ask counsel what the policy should be. The order we recommend is consent, signals, masking, retention, deletion, paperwork. None of it is legal advice; have counsel review your setup before you turn recording on.
Start with a lawful basis, not with the recorder
Article 6 of the GDPR requires a lawful basis before the first event is stored. For replay, teams usually pick one of two: consent for product research, or legitimate interests for security and debugging, supported by a documented balancing test and a recording limited to that purpose. If one library serves fraud review and conversion research, that is two purposes; name both.
Consent gates the recorder, not the banner
Consent has to sit in front of the recorder, not next to it. If the visitor has not opted in to your replay category, the SDK never loads and no events are produced. A banner over a running recording is a disclosure, not consent. Withdrawal needs to be as easy as acceptance: when consent is revoked, stop recording and delete that visitor’s stored sessions.
Honor Global Privacy Control before the first event
Global Privacy Control is a browser-level opt-out preference. California regulations require businesses to honor it, and honoring it everywhere else is a defensible default. Implementation is boring on purpose: read navigator.globalPrivacyControl at page load, and if it returns true, do not start the recorder, do not persist a session, drop queued events, and log the decision with a timestamp. The usual failure is honoring the signal in the cookie banner and ignoring it inside the replay SDK. If you rely on legitimate interests, the same check also covers an objection under Article 21 — one gate, two purposes.

Mask in the browser, then mask again on the server
Masking decides whether your replay library is an asset or a liability. In the browser, enable all-input masking in the recorder and add selectors for text that carries personal data outside inputs. On the server, re-mask known sensitive fields before the session is written to storage: client-side masking depends on code you control today, and DOM changes, iframes and third-party widgets can slip past it. Prynt Replay masks every form field in both places, with no unmask option — a reveal toggle would turn a masked recording into a personal-data incident during a support call.
What masking has to cover
- Password, card, tax ID, phone and email inputs.
- Free-text areas, where people paste addresses and order numbers.
- Text outside inputs: confirmation screens, account headers and chat widgets.
- Query strings and request payloads captured alongside the DOM.
Then test it: script a signup with fake data, open the replay and search the stored payload for the fake string. If you can find it, that field was not masked. Pair the recorder with form-level protection so the fields you mask are the ones you validate.
Set retention by purpose and enforce it in code
Article 5(1)(e) asks you to keep personal data no longer than the purpose requires, and replay rarely needs a long window: rage clicks, JavaScript errors and checkout drop-off surface within days, not months. Choose a number, publish it in your privacy notice and enforce it in the system, not in a spreadsheet. Prynt Replay retains sessions for 30 days. If your purpose is narrower — debugging one release — shorten the window and export only the sessions you investigated. If replay doubles as fraud evidence, keep that copy in a separate store with its own retention rather than extending the whole library.

Build deletion for one person, not a batch job
An erasure request under Article 17 arrives one at a time and names a person, not a session identifier. Your deletion path therefore starts from something you can search: an email address, an account ID, or a visitorId from the identity layer. Sharing that identifier between replay, device fingerprinting and your decisions is what makes the search one query instead of five.
Then run it: resolve the person to every identifier you hold; find sessions by those identifiers, not by date range; delete the recordings and any derived aggregates that still point to an individual; log the timestamp and scope without keeping the deleted content; confirm to the requester once you have verified. Read the processor terms too — ours are on our DPA page.
Have counsel review the setup, then review it again
Put a review on the calendar with your counsel or DPO: the lawful basis per purpose, the consent gate, the Global Privacy Control check, the masking test result, the retention number and the deletion runbook. Keep the artifacts that prove each one; they answer a regulator faster than a paragraph of policy.

FAQ
Is session replay allowed under GDPR?
Yes, when you have a lawful basis, collect only what the purpose needs, limit retention and can erase on request. The hard part is evidence, not permission.
Does masking remove the need for consent?
No. Masking reduces personal data in a recording; it does not create a lawful basis. You still need consent or a documented legitimate interest — and you still need to honor opt-outs.
How long should we keep session replays?
As short as your purpose allows. Prynt Replay retains sessions for 30 days; if you record for release debugging, shorten that and export only what you investigate.
Try it free
Prynt Replay masks every form field by default, keeps sessions for 30 days and costs $49/month after a 14-day free trial: pryntid.com.
Try it free
Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.