Abstract Aims Since 2008 England's anti-stigma programme Time to Change has lobbied media outlets about stigmatising coverage and worked with them to promote accurate and non-stigmatising coverage. While this may have an impact on coverage and hence attitudes, it is also possible that coverage can change in response to improving attitudes, through the creation of a market demand for less stigmatising coverage. This study evaluates English newspaper coverage of mental health topics between 2008 and 2016. Method Articles covering mental health in 27 newspapers were retrieved using keyword searches on two randomly chosen days each month in 2008–2016, excluding 2012 and 2015 due to restricted resources. Content analysis used a structured coding framework. Univariate logistic regression models were used to estimate the odds of each hypothesised element occurring in 2016 compared with 2008 and Wald tests to assess the overall statistical significance of the year variable as the predictor. Results The sample retrieved almost doubled between 2008 (n = 882) and 2016 (n = 1738). We found a significant increase in the proportion of anti-stigmatising articles (odds ratio (OR) 2.26 (95% confidence interval (CI) 1.86–2.74)) and a significant decrease in stigmatising articles (OR 0.62 (95% CI 0.51–0.75)). Reports on all diagnoses except for schizophrenia were more often anti-stigmatising than stigmatising. Conclusions This is the first clear evidence of improvement in coverage since the start of Time to Change. However, coverage of schizophrenia may be less affected by this positive shift than that of other diagnoses. The increase in the level of coverage identified in 2016 requires further investigation, as it may also influence public conceptualisation of what constitutes mental illness, attitudes to mental illness in general and/or specific diagnoses. While most anti-stigma programmes are not diagnosis specific, we suggest their evaluation would benefit from a diagnosis specific approach to allow fuller interpretation of their effects. This could include media analysis driven by hypotheses based on diagnoses to ascertain whether variations by diagnosis over time occur both in the nature and in the proportion of coverage.