The hazard perception section of the UK theory test trips up more candidates than the multiple choice — and the reason is almost never that people don't know what a hazard looks like. The reason is that they don't understand how the scoring works, and they practise in the wrong way. The score you need is 44 out of 75. Most candidates who fail the hazard perception section have the knowledge to pass it. What they lack is calibration.
What hazard perception is actually testing
The test is not asking whether you can spot objects in a video. It is asking whether you can recognise a situation as it starts to become dangerous — before it is fully dangerous. That distinction is the whole thing. The DVSA calls this the difference between a potential hazard and a developing hazard. A parked car with a driver visible inside is a potential hazard — it could pull out. The moment it begins to indicate or edge forward, it has become a developing hazard. A cyclist ahead on the road is a potential hazard. The moment they swerve toward the centre of the road to avoid a drain, they become a developing hazard.
Under test conditions, the natural instinct is to wait until you are certain — to click only when the danger is unambiguous. By that point, the highest-scoring part of the window for that hazard has often already passed. The test rewards the recognition of danger starting to unfold, not confirmation that it has fully unfolded.
The 2024 CGI format change
From April 2024, the DVSA replaced the live-action video clips used in the hazard perception test with CGI (computer-generated imagery) animated clips. This change was made to allow more consistent and controllable test content. The CGI clips show driving scenarios in the same environments as the old footage — urban streets, rural roads, dual carriageways — but in a rendered format.
The scoring principles are identical and the skill being tested is the same. What changes is the visual character of the footage: the CGI clips have a slightly different look to real video and candidates who have only practised with pre-2024 live-action material may find them initially unfamiliar. Most current revision apps and the DVSA's own practice materials have been updated to use CGI-format clips. Confirm any revision app you use has updated its clip library if you purchased it before 2024.
How the scoring actually works
Each clip has a scoring window that opens at the moment the hazard begins to develop. Click at the very start of the window and you score 5 points. Click progressively later in the window and you score 4, 3, 2, or 1 point depending on how late you respond. Click after the window has closed — even slightly after — and you score zero for that hazard. You are not told where the window opens or closes. This is deliberate: the DVSA does not want candidates memorising timecodes from practice clips and clicking at the exact right moment on autopilot. The test is measuring genuine hazard recognition, not pattern memory.
The practical implication is that practising with the same clips repeatedly until you can predict when to click is the wrong approach. The right approach is to practise with a large variety of clips until clicking earlier becomes your instinctive response to any developing situation.
The 14 clips and the double hazard
You see 14 video clips in total during the test. Thirteen contain one scoreable developing hazard each. One clip contains two scoreable developing hazards — and you need to respond to both to maximise your score on that clip. This double-hazard clip is not announced or signposted. A significant number of candidates who score well on the other 13 clips drop unnecessary points on the double-hazard clip because after clicking the first hazard, the instinct is to relax and wait for the clip to end. Stay alert throughout every clip from its first frame to its last, regardless of whether you've already clicked a hazard within it.
Why clicking rapidly zeroes your clip
The DVSA's scoring software monitors click patterns throughout each clip. If it detects a pattern of rapid, repeated clicking within a short period — the threshold is not published, but clicking more than roughly once per second in a burst is the commonly reported trigger — it scores that clip as zero automatically, regardless of whether any of those clicks would otherwise have fallen within the scoring window. The system interprets rapid clicking as an attempt to game the test by saturating the clip with clicks.
The right approach is to watch each clip as though you are actually driving the road. When you see something beginning to develop — a pedestrian stepping toward the kerb edge, a cyclist who appears to be moving position, a junction ahead where a car is clearly preparing to emerge — click once. If the situation continues to develop further, a second click is fine. Then stop and watch. Considered clicking, not clicking to maximise coverage.
The hazard types that trip candidates most often
Junctions with emerging vehicles are the most common single hazard type and the one where late clicking is most penalised. The hazard begins the moment the other vehicle starts moving toward the junction, not when it reaches it. Pedestrians crossing or beginning to step off the pavement are similarly early-scoring — the click should come as they begin to move into the road, not as they reach the middle of it.
Vulnerable road users — cyclists, motorcyclists, and pedestrians — appear frequently and score zero for candidates who click only when the conflict is inevitable rather than developing. A cyclist near a parked car is a potential hazard. A cyclist moving away from the kerb to pass that parked car is a developing hazard the moment their line changes. Practising specifically with clips that feature cyclists and pedestrians is worth doing as a targeted exercise if mock scores show consistent misses on these scenarios.
Common mistakes and how to fix them
The most common failure mode is clicking too late. The natural response when watching a clip is to wait for confirmation that the hazard is real before committing to a click. By the time the hazard is undeniable, the highest-scoring zone has often closed. The fix is deliberate practice with a conscious intention to click earlier than feels comfortable — and then reviewing the score to see whether early clicking produced better results. It almost always does.
Missing the double-hazard clip is the second most common source of dropped points. If you click one hazard and the clip continues — particularly if there is still developing action on screen — look for the second hazard rather than assuming the clip is winding down.
Treating hazard perception as the easy section and not practising it is the third mistake. The pass rate data does not support this assumption. Hazard perception is a perceptual and reflexive skill, not a knowledge-based one. Reading about it, including reading this article, will not move your score as much as watching actual practice clips and clicking in real time over several sessions. Build clip practice into your revision from the beginning, not as a last-day addition.
Scoring benchmarks and practice targets
The pass mark is 44 out of 75. A score of 55 or above on practice clips suggests genuinely solid calibration. If you are consistently scoring below 44 after two weeks of daily clip practice, the problem is almost always one of two things: clicking too late on everything, or consistently missing a specific scenario type. Both are diagnosable. Review which clips produce zero scores and identify whether there's a pattern — emerging vehicles at junctions, cyclists, pedestrians — and target that type specifically in subsequent practice sessions. Isolated zero scores on random clips are usually late-click calibration. Consistent zeros on one type of scenario point to a recognition gap that targeted practice closes quickly.
For more theory test prep, see our guide on passing the multiple choice section first time. Once you have your licence, browse used cars under £5,000 on AllCarsUK.