A writers' union's own survey ties AI use to falling income

The writer's problem

A UK author trying to gauge whether generative AI is actually affecting working writers' income, rather than just their anxieties, needs a source willing to publish numbers rather than impressions. The Society of Authors sets out its position and evidence on its advocacy page on generative AI, copyright and freedom of expression.

What the documents show

The SoA states that in its own membership surveys, 86 percent of authors say generative AI has already reduced their earnings, 72 percent say job opportunities have already been cut, and 57 percent say their career is no longer sustainable because of it; a jointly published report the SoA co-launched with four other creator organizations, Brave New World? Justice for Creators in an age of Gen AI, dated 30 January 2026, states these figures draw on responses from over 10,000 creators across the five organizations. These are self-reported findings from a self-selected group of respondents to an advocacy body's own survey, not a randomized study of the writing profession, a distinction the SoA's page does not itself make explicit. Separately, the SoA's Human Authored scheme lets members who sign an agreement add a digital logo to books they say were written without generative AI, a self-declaration rather than a third-party audit.

The editorial choice

The SoA uses these figures to argue for a specific policy position, a licensing-based framework it calls CLEAR, rather than treating the numbers as a neutral research finding; a reader citing this survey should carry that advocacy framing forward rather than presenting the percentages as an independent academic result.

What stays with the author

Deciding whether to display a Human Authored logo, and standing behind that declaration, remains each author's own choice and responsibility; the scheme itself does not verify the claim beyond the signed agreement. The SoA's own data does not establish causation between generative AI specifically and any individual author's income change, only members' self-reported attribution of it.

• Is a cited AI-income figure from a controlled study, or from an advocacy organization's own member survey?
• Would a Human Authored declaration hold up if a reader asked exactly what 'without generative AI' excluded?
• What does the underlying report actually say about its response rate and selection method?

The Society of Authors' guidance is a working writers' organization's own account of its members' experience, useful precisely because it says so, not because it claims to be something broader.