Can anxiety therapy make you feel worse before it helps?

Exposure-based work can raise anxiety in the short term, because it deliberately approaches avoided situations. What matters is the direction of the trend over several weeks, not one hard session. A score that keeps rising counts as reliable deterioration in NHS Talking Therapies reporting, so tell your therapist promptly if that is happening.

What if my GAD-7 score jumps around from week to week?

Single-week wobbles are common, which is why a reliable change index exists. The NHS manual sets that index at 4 points or more on the GAD-7, so smaller moves may be measurement noise. Look at the line across four to six sessions instead of comparing two adjacent weeks.

Do app-based or online CBT programmes track progress the same way?

Many do, and the research base supports it. A meta-analysis found the alliance-outcome correlation for internet-based psychotherapy was r = 0.275 across 23 studies, close to the face-to-face figure. NICE also describes guided self-help for GAD as 5 to 7 supported sessions with review of progress and outcome.

Should I stay with a therapist I like if my scores are not moving?

Liking your therapist is not a reason to ignore a flat graph, and a flat graph is not automatically a reason to leave. NIMH suggests raising it directly with the therapist first, then considering other professionals or approaches. A GAD meta-analysis found that adding more sessions of the same thing was unrelated to the size of the effect.

How long should therapy continue after I start feeling better?

NICE recommends that a course of treatment includes relapse prevention work, which can continue past the point of recovery. The NHS manual also says treatment may need to continue when disability scores stay high even after symptom scores drop. Discuss a planned ending rather than stopping the week you first feel well.

How Do You Know If Anxiety Therapy Is Working?

The short answer: Progress in anxiety therapy is measurable, not just a feeling. Most services score a short questionnaire at every session, usually the GAD-7. A fall of 4 or more points on the GAD-7 is the level England's national talking therapies programme counts as reliable improvement. Change by about week four of a course predicts the final outcome, so a flat score after a month of weekly sessions is a reason to review the plan.

How soon should anxiety therapy start to work?

Evidence-based courses are shorter than many people expect. The NICE guideline on generalised anxiety disorder and panic disorder says CBT for GAD usually runs 12 to 15 weekly sessions of 1 hour, fewer if the person recovers sooner. For panic disorder the same guideline sets an optimal range of 7 to 14 hours in total, in weekly sessions of 1 to 2 hours, completed within a maximum of 4 months. Guided self-help for GAD is shorter still, usually 5 to 7 sessions of 20 to 30 minutes.

A systematic review of 26 dose-response studies in routine services found optimal doses between 4 and 26 sessions, and 4 to 6 sessions for low-intensity guided self-help. The same review found weekly therapy speeds up the rate of improvement compared with less frequent appointments. So the first checkpoint is not month six. It sits closer to week four.

That checkpoint has direct evidence behind it. In a study of internet-delivered CBT for depression, social anxiety disorder and panic disorder, change at week four was the best predictor of final outcome. A disorder-specific measure at week four explained 41% of the outcome for depression, 43% for social anxiety disorder and 34% for panic disorder. General mental health measures predicted much less.

Which scores tell you if therapy is working?

The default anxiety measure is the GAD-7, a seven-item scale developed in 15 US primary care clinics with 2,740 adults who completed the study questionnaire. At its optimal cut point the scale reached 89% sensitivity and 82% specificity against interviews by mental health professionals. Rising scores tracked worse function on all six scales of a standard health survey, plus more disability days.

Good services measure more than symptoms. The NHS Talking Therapies manual requires the PHQ-9 for depression, the GAD-7 for anxiety and the Work and Social Adjustment Scale at every session. It also requires a disorder-specific measure where one applies: the Social Phobia Inventory, the Panic Disorder Severity Scale, the Obsessive-Compulsive Inventory, the PTSD Checklist for DSM-5 or the Health Anxiety Inventory.

That detail changes results. Analyses of the national database showed better clinical outcomes, and larger falls in mental health related disability, when therapists used the relevant disorder-specific measure for PTSD, social anxiety disorder, panic disorder and OCD. Depression scores belong in the picture, because the two problems often travel together, and depression.md covers how the PHQ-9 side is read.

How big a score change counts as real progress?

Two numbers do most of the work. The NHS manual sets GAD-7 caseness at a score of 8 or above, and the reliable change index at 4 points or more. A separate study of 261 patients estimated the minimal clinically important difference for the GAD-7 at 4 points. The two thresholds agree.

The same manual defines the three outcomes services report:

Recovery: you started treatment above the clinical threshold and finished below it.
Reliable improvement: your score fell by a reliable amount, and neither measure rose reliably.
Reliable recovery: you meet both of the conditions above.

Thresholds differ by measure. Caseness is 32 or above on the PTSD Checklist for DSM-5 and 19 or above on the Social Phobia Inventory, with reliable change set at 10 points for each. The programme aims for 53% reliable recovery and 71% reliable improvement by 2028/29. Those are service averages, not a prediction for one person. Note also that the 4-point study ran inside a chronic depression trial, and its authors asked for confirmation in anxiety-specific samples.

What progress do symptom scores miss?

Symptom relief and daily function are not the same thing. The Work and Social Adjustment Scale rates how much a problem interferes with work, home management, social life, private leisure and close relationships. Its validation study reported a test-retest correlation of 0.73, and correlations of 0.76 with depression severity and 0.61 with obsessive-compulsive symptom severity.

The NHS manual is blunt about the gap. Disability does not always fall as symptoms fall, so clinicians are told to watch the disability score as well. If it still shows real interference after the GAD-7 and PHQ-9 have dropped, further sessions focused on functioning may be needed. The manual also warns that people with marked avoidance, such as agoraphobia, can score below caseness on the GAD-7 while being severely disabled.

Functional markers worth tracking yourself:

Situations you now enter that you avoided a month ago.
Safety behaviours you have dropped, such as sitting near exits or repeated checking.
Work, study or social commitments you keep without a long build-up.
How fast you recover after a bad day, rather than whether bad days still happen.

Practice between sessions drives much of that change. Our guide to anxiety management techniques that work covers the homework side.

Does the relationship with your therapist predict results?

It correlates with them. A meta-analysis of 295 studies and more than 30,000 patients found an alliance-outcome correlation of r = 0.278 for face-to-face psychotherapy, equivalent to d = 0.579. For internet-based therapy the figure was almost the same, r = 0.275 across 23 studies. The link held across rater perspective, measure, treatment approach and country. The authors still treat causal direction as an open question, so a strong alliance is a signal, not a guarantee.

Seeing your own scores appears to help. A meta-analysis of 24 routine outcome monitoring studies found two-thirds showed monitored therapy beating usual care from the same clinicians. Feedback reduced deterioration rates, and nearly doubled rates of clinically significant or reliable change among patients predicted to do poorly. Asking to see your score graph is a fair request. NIMH suggests asking at the start how progress will be assessed, and whether a time frame or session count is planned.

When should you switch therapist or add medication?

Guidance sets decision points rather than one fixed deadline. Under NICE, if GAD has not responded to a full course of a high-intensity psychological therapy, drug treatment should be offered. If it has not responded to drug treatment, the options are a high-intensity psychological therapy or an alternative drug. A partial response to a drug is a reason to consider adding psychological therapy rather than waiting longer.

For panic disorder, NICE says that once two interventions have been provided in any combination and significant symptoms remain, referral to specialist mental health services should follow. Medication carries its own review schedule: efficacy and side effects reviewed within 2 weeks of starting, again at 4, 6 and 12 weeks, then a decision at 12 weeks on whether to continue or change. Drug timelines differ from therapy timelines, as covered in how long anxiety medication takes to work.

More sessions alone may not fix a stalled course. A meta-analysis of 41 randomised trials in adults with GAD found therapy duration and number of sessions were unrelated to treatment effect. Sessions lasting longer than 60 minutes produced larger effects than shorter ones (b = 0.62, 95% CI 0.21 to 1.05). So if scores are flat, the content and format of sessions deserve as much scrutiny as the total. Raise it with your therapist, and see a doctor about anxiety if nobody is tracking your scores at all.

The bottom line

Ask for a score at every session, then read the trend instead of one week. A 4-point fall on the GAD-7 is the accepted marker of reliable change, and a final score under 8 is the usual recovery threshold. Track function as well as symptoms, because avoidance can hide behind a low symptom score. If nothing has shifted by roughly week four of a weekly course, say so and ask what changes next.

Last updated: September 2026. This article is for informational purposes only and does not constitute medical advice. Do not start, stop or change anxiety treatment, including therapy or medication, without speaking to a qualified clinician.