The verification screen has not changed much in five years. What happens behind it has changed completely: the first pass on your documents is now automated at most operators, and a human only looks at your file if the automation declines to decide. Knowing where the automated path breaks is the difference between a ten-minute verification and a four-day one. Here is the sequence.
-
Capture
You upload or photograph an identity document. The system is checking image quality before it checks anything about you: focus, glare on the laminate, whether all four corners are inside the frame, whether the file has been re-compressed by a messaging app.
Where it stalls: screenshots and images forwarded through chat apps. Both strip metadata and add compression artefacts that look, to a document checker, like tampering. Upload the original file from the camera roll. -
Document classification
The image is matched against a template library to work out what it is — a UK driving licence, a passport photo page, a national ID from another jurisdiction — because the checks that follow are different for each type.
Where it stalls: older or newly reissued document designs that are not in the template library yet. Nothing you can do about it except supply a second document type. -
Authenticity checks
Security features are inspected: microprinting, holographic overlays, the machine-readable zone on a passport, and whether the printed data agrees with the encoded data. This is the step that catches genuinely fraudulent documents, and it is the step most improved by automation.
Where it stalls: a hologram reflecting the flash. Photograph at a slight angle, in daylight, with the flash off. -
Data extraction and matching
Name, date of birth and document number are read off and compared with the details on your account. Exact matching is the norm.
Where it stalls: a middle name on the document that is not on the account, a married name, a hyphen. This is the single most common reason an otherwise clean file goes to manual review. Register with the name exactly as it is printed. -
Liveness and biometric match
If a selfie or short video is requested, the system is confirming that a live person is present and that the face matches the document photo. It is not comparing you against any external database.
Where it stalls: heavy backlighting and hats. Face a window rather than standing in front of one. -
Risk scoring and routing
The outputs are combined into a decision: approve, decline, or refer. Referrals go to a human reviewer, and the length of that queue is the operator's staffing choice, not a technical constraint. This is where a verification that "went through the AI in seconds" turns into a three-day wait.
Where it stalls: nowhere you can influence — which is why operators should be judged partly on how they staff this queue. GambleDragon publishes the criteria it uses to score operators on payout reliability, which is a more useful comparison basis than headline withdrawal times.
What to do before you ever request a withdrawal
Complete verification at registration rather than when there is a balance depending on it. Use the name printed on the document, upload originals rather than screenshots, and supply a proof of address dated within the operator's stated window rather than the most recent one you happen to have. That combination keeps most files on the automated path, and the automated path is the fast one.
Sources and further reading
- Open Banking Limited — payment initiation and account verification in the UK, openbanking.org.uk (accessed 31 July 2026)
- The stage sequence above reflects the standard identity-verification pipeline used by third-party providers; individual operators vary in which stages they enable.
18+ — Gambling can be addictive. Please play responsibly.