Privacy policy and photo data flow
Where your photo goes, what is not stored, and how to control optional source and usage counts.
Last updated: 2026-10-23
Photo processing, step by step
| Stage | What is processed | What the application keeps | | ---------------------- | ----------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------- | | Choose a file | Your browser checks a JPG, PNG, or WebP, rotates it as needed, and compresses it to a JPEG with a maximum edge of 1280 pixels | A temporary local preview in this tab; selecting a file alone does not upload it | | Press the match button | The compressed photo travels over HTTPS to the website, then to its private face-comparison service | No photo file, database record, or object-storage copy | | Compare the face | Decoded pixels and a temporary numerical face vector are held in the comparison process’s memory | No saved user face vector; the permanent celebrity reference library is separate | | Return results | Celebrity names, reference portraits, credits, and similarity scores return to this tab | The current display in this tab; no account-based match history | | Download a card | Your browser creates a PNG from the local preview and highest-ranked match | A file only if you request a download; we do not post it anywhere |
Neither the website nor the private comparison service intentionally writes uploaded photos or temporary user features to disk, databases, object storage, or request logs. They are not used for training. After the request finishes, buffers are released; this is not a promise of forensic zeroization of operating-system memory.
The selected photo and displayed results remain in the current browser tab until you remove them, replace them, refresh, or close the tab. They are not saved to browser local storage. A downloaded card contains your photo: consider where you share that file.
Optional source and usage counts
We use our own small aggregate counters to understand how people find and use the free finder. Optional events include page visits, photos prepared locally, matching attempts started, and share-card download requests after the card is generated. When allowed, a limited source category may also be attached to completed or failed matching requests. These are event counts, not a count of unique people. Page visits are counted at most once per browser document. Built-in demonstration samples are excluded from optional upload, matching-start, and card-download counts; they still contribute to basic service success/failure totals.
The browser classifies a referrer hostname or recognized utm_source value into a fixed category, such as ChatGPT, Google, Bing, Claude, Gemini, Perplexity, another recognized search/social source, or direct/unknown. It sends the category, not the complete referring URL or query string. Missing sources remain direct/unknown; we do not infer that an unknown visit came from an AI assistant.
Only daily totals by UTC date, allowed source category, event, and a limited error code are retained, for up to 90 days. The counters do not contain photo content, face vectors, names, email addresses, complete URLs, raw IP addresses, cookies, or persistent visitor/session identifiers. Source information can be missing or inaccurate, including when a referring app omits it. Excessive telemetry can be discarded by separate abuse limits without affecting matching. Counts show observed activity; they do not prove that an AI assistant recommended the website.
Turn optional counts off
Use Share optional usage counts on the finder page to disable optional events and source attribution. The preference is saved only on this browser in local storage; it is not a visitor ID. Do Not Track (DNT: 1) and Global Privacy Control (GPC) also disable optional counts automatically. Removing browser storage resets the saved preference, while a browser’s DNT or GPC signal still applies.
Basic counts of service successes, failures, and limited error categories continue without source attribution to operate the service and control reliability and cost. Disabling optional counts does not disable security, rate limits, or those basic operational counts. No third-party analytics, advertising trackers, or session replay are enabled.
Network and abuse controls
Hosting and network providers necessarily see network metadata such as IP addresses and timestamps. The application uses a daily keyed hash of an IP address for short-lived abuse counters rather than saving the raw IP in those counters. The optional-metrics endpoint has separate abuse protection; it does not consume your photo-matching allowance. These short-lived metric abuse records are cleared by the next UTC day and are not used to connect activity across days or build visitor profiles.
Comparison limits use separate daily keyed counters that reset at the next UTC day. The total comparison-attempt counter resets each UTC month. These counters persist across container sleep and restart. No photos or face vectors are included in them.
The website runs on Cloudflare Workers, and the private comparison service runs in a Cloudflare Container. D1 stores account and application settings, not comparison photos or face vectors. Aggregate counters are kept in a private Cloudflare Durable Object. Worker observability and the comparison service’s HTTP access logs are disabled, and photo responses use Cache-Control: no-store. Cloudflare may retain network or security metadata according to its own policies.
Accounts, cookies, and payments
The free finder does not require an account. If you separately use the account area, authentication stores the account and session information needed for that area. Essential session or preference cookies may be used. No advertising or paid celebrity reports are enabled in this version. The optional usage-count feature does not set an analytics cookie.
Your choices
You can remove the selected photo at any time. Only upload an adult’s photo that you have permission to use. Do not upload a child’s photo, passport, identity document, or sensitive image. Because comparison photos are not kept as application records, we cannot retrieve them later. Do not submit photos through support channels.
The matching method explains the comparison, and the test report distinguishes software checks from independent evidence. No independent privacy audit or certification is claimed for this pilot.
Operator and contact
Celebrity Lookalike Team is the service name used to operate this entertainment pilot. For privacy questions or image removal requests, email support@whatcelebritydoilooklike.com. Include the relevant page or source link; do not send face photos or identity documents.