When AI reads your Annual Report first, what’s left of your story?
Most of the AI conversation in reporting has been about whether AI can find your content. Our 6 October webinar looked at the next question: what happens to your story once it does?
Stephen Butler, our Reporting & Advisory Director, was joined by Luna Raphael, Luminous’s Creative Director and Digital Lead. Also on the panel were Hannah Boore, Senior Manager, Investor Relations at Lloyds Banking Group, and Ross Hawley, Head of IR at ZIGUP and Chair of the IR Society’s AI Steering Group. Here’s what we learned.
Accurate isn’t the same as distinctive
Over the summer we asked AI tools the questions investors ask, then compared the answers with companies’ own reporting. AI mostly got the facts right, but the stories came back flatter. On average, companies lost 16% of their distinctiveness. Business models suffered most. In one review, AI skipped the business model altogether and went elsewhere to work out how the company made money.
The reports that held up best followed a simple pattern: a clear claim, how it differs from peers, the proof, and the “so what”. That last part matters most, because AI tends to keep the metric and drop what it means.
Get your own content into the answer
When Hannah asked AI tools about Lloyds’ key messages, none of the sources were Lloyds. They were Reddit threads, news articles and blogs. So the team moved key content from the PDF into HTML. They added searchable performance tables and AI-friendly FAQs, and they used plainer wording. The result was a 71% rise in AI-referred traffic year on year, and the site is now cited around five times more than the PDF. They’ve since done the same for results day, which Ross pointed out is the moment that actually moves the share price.
Clean up the data
Ross cleaned up the metadata on 300 documents. He made the company the author, added proper descriptions and fixed the keywords, and his discoverability score went to 20 out of 20. Luna added the technical basics: pages that work without JavaScript, CDN settings that don’t block AI crawlers, and a text version of every chart and video. Her rule of thumb: “Design and structure your pages with people in mind, build them so the machines can read them.”
Use AI to think, not to write
The panel agreed AI earns its place in proofreading, consistency checks, condensing copy and sentiment analysis. First drafts are a different matter. In Ross’s words: “Don’t delegate your drafting to AI. Use AI to give yourself the space to think and make the right judgement calls.”
One thing to do this week
Hannah’s advice: put your key messages, and the questions investors ask, into AI tools. See what comes back, what gets lost, and whether the things that set you apart survive.
Watch the full recording below:
Want to see how your investor story holds up when AI retells it? Get in touch.
