Oxford researcher: AI is a worryingly good persuader, but its real-world impact is overrated
Felix Simon, an Oxford researcher studying AI's social impact and FT tech commentator, published a series of articles in Transformer on September 2 analyzing AI persuasion and its real-world effects; Transformer officially amplified the piece the next day. His core conclusion: AI is already a worryingly good persuader, but its real-world influence tends to be overrated, because the bottleneck is not persuasion but reach.
Confirmed
- Striking experimental evidence: a preprint study found chatbots could even outperform seasoned activists at getting people to hand over money—unthinkable five years ago; multiple studies confirm AI is a worryingly good persuader.
- Simon distinguishes persuasion, conducted through verifiable arguments, from manipulation.
- His core argument: persuasion is the easy part; the hard part is capturing attention, which is scarcer than persuasion in an information-saturated environment.
- In a follow-up post he adds: experiments differ from the real world; people are constantly exposed to contradictory information; individuals vary in persuadability and can resist persuasion attempts—together with limited reach, these complete the case that AI's persuasive impact is overrated.
- He also recommends Science journalist Kai Kupferschmidt's long-form report on the mechanisms of AI persuasion, crediting the Transformer team and scholar Ben Tappin; the report's key reminder is that real-world impact is far milder than experimental data suggest.
Why it matters
- The analysis cools fears that persuasive AI will upend public opinion or manipulate the masses: persuasive capability does not equal real-world influence; reach and attention allocation are the key bottlenecks.
- For policy and research, attention may need to shift from whether AI can persuade to propagation-chain questions such as whether AI can be seen and trusted.
2026-09-02 ~ 2026-09-03 · 6 related posts
Primary sources
- AI is a scary-good persuader, but its real-world influence will be smaller than feared: Oxford researcher — _FelixSimon_ ·
- Felix Simon adds three more reasons AI persuasion won't dominate the real world — _FelixSimon_ ·
- Science deep dive examines how AI persuasion works — and why real-world effects may be milder — _FelixSimon_ ·
- [source] Science deep dive examines how AI persuasion works — and why real-world effects may be milder — _FelixSimon_ · 2026-09-02
- [source] AI is a scary-good persuader, but its real-world influence will be smaller than feared: Oxford researcher — _FelixSimon_ · 2026-09-02
- [source] Felix Simon adds three more reasons AI persuasion won't dominate the real world — _FelixSimon_ · 2026-09-02
- Persuasion Is the Easy Part: Why Persuasive AI's Real-World Impact May Be Overhyped — _FelixSimon_ · 2026-09-02
2 near-duplicate retellings: _FelixSimon_ · ShakeelHashim