ERP has strong evidence, which means that when it isn’t working the useful question isn’t whether the treatment works but what is different about how it’s being delivered. Four causes account for nearly all of it.
1. Mental compulsions are still running
By a wide margin the commonest. You do the exposure, prevent every visible compulsion, and ruminate quietly throughout — arguing, reassuring yourself, reviewing, checking whether the anxiety has dropped. The neutralising happens internally, so the new learning never occurs.
Signs: exposures that produce distress but never get easier; a sense of getting through them by white knuckling; anxiety that drops sharply the moment the exposure ends. If this is it, the fix is to make the covert compulsions explicit targets, which usually requires a clinician who knows how to look for them.
2. Safety behaviors are attached
The exposure is being done with something that takes the edge off. Only when someone is home. Only with your phone in hand. Quickly, to get it over with. Distracted. While silently repeating that it’s only OCD.
Each one signals to your nervous system that a precaution was necessary, which is the opposite of the intended lesson. Response prevention has to be complete for the item you’re working on.
3. Not enough, or not often enough
ERP is mostly homework. Weekly sessions with no daily practice won’t move a disorder that has been reinforced thousands of times. Similarly, exposures pitched too low — staying at a 20 out of 100 because the 50 is frightening — feel like progress and teach little.
Sometimes the answer is simply intensity:an intensive programme compresses months of weekly work into weeks and gets people moving who had stalled.
4. Something else is in the way
Severe depression makes exposure work extremely difficult and may need treating first. Untreated ADHD makes daily homework structurally hard. Active substance use interferes. Significant trauma may need to be addressed alongside. And occasionally the OCD is severe enough that medication is needed before therapy can gain traction.
The fifth possibility, said plainly
The clinician may not be trained well enough. If they never ask what you’re doing in your head during exposures, if they reassure you when you’re distressed, if sessions have drifted into discussing the content of your obsessions, or if there’s no hierarchy and no homework — you may be receiving something other than ERP.
The five screening questions apply to a clinician you already have, not only to a new one.
Before you conclude it failed
Give it a proper trial: twelve to twenty sessions with daily homework, with mental compulsions explicitly targeted. If that has genuinely happened and there’s no movement, the next options are an intensive programme, medication optimisation, or a different modality —I-CBT in particular has evidence for people who couldn’t tolerate exposure.