Background: As many sufferers from phobic and panic (phobia/panic) disorders cannot get to suitable therapists, routine aspects of therapy were delegated to internet-accessed computer-aided self-help with or without exposure instructions. Methods: Phobia/panic referrals were randomised to computer-aided self-help via the internet at home in a 2:1 ratio either by self-exposure cognitive behaviour therapy (CBT) [FearFighter (FF), n = 45] or by minimal CBT without exposure [Managing Anxiety (MA), n = 23]. All had brief backup phone advice from a clinician concerning their computer guidance. Results: On self-ratings and blinded assessor ratings, patients improved equally with each form of self-help over 10 treatment weeks but significantly more on 5 out of 10 measures by week 14 (1-month follow-up) when the self-help included self-exposure instructions than when it did not. In accord with this, standardised effect sizes (Cohen’s d) indicated superiority of FF over MA on 5 measures by week 14. Satisfaction with treatment in all patients pooled correlated positively with improvement after treatment and at 1-month follow-up. Conclusions: At the end of treatment, computer-aided CBT self-help at home via the internet plus brief live helpline support was effective with or without exposure instructions, and at 1-month follow-up it was more effective on some measures if exposure instructions had been included. Analysis is needed of how non-exposure CBT produced its shorter-term effect.