Explore every episode of the podcast Beneath the Surface
| Title | Pub. Date | Duration | |
|---|---|---|---|
| ChatGPT cites Russian propaganda. It can also swing elections | Beneath the Surface Ep.2 | 13 Jul 2026 | 00:31:57 | |
Ask ChatGPT about the war in Ukraine, and there's a one-in-five chance it answers with Russian propaganda. This episode is about how that happens – and why it matters far beyond Ukraine. Julia Petryk runs Calibrated, a PR agency in Kyiv, and has lived and worked in Ukraine since the start of the war: sleepless nights under Russian missiles and drones, blackouts so total she dug a reflective scarf out of her wardrobe just to cross the road safely, a winter at minus fifteen with no heating after Russia hit her neighbourhood's plant. She knows better than almost anyone that Russia's war isn't just fought with missiles – it's fought with information. Julia describes what that looks like from the inside: AI as an accelerator of propaganda, the Doppelganger campaign that cloned real news websites to pump out fake stories, and an uneven fight where Russia spends billions on disinformation while Europe spends millions to counter it. And that propaganda is turning up somewhere you might not expect: inside your chatbot. Research from the Institute for Strategic Dialogue found that when you ask ChatGPT, DeepSeek and Grok about the war, almost a fifth of responses cite Russian state-attributed sources. We break down – in plain language – how chatbots actually source their answers, the difference between the base model and a live search, and why "garbage in, garbage out" means Vladimir Putin has basically ended up doing GEO. Then the question that really matters: can a chatbot actually change your mind – even about how you vote? We speak to Hause Lin, a research scientist at MIT who ran one of the largest studies of its kind: around 2,400 US voters, each having a roughly six-minute conversation with a chatbot arguing for Trump or Harris right before the 2024 election. The result was a 2.88 percentage-point shift – one in 27 people switched their vote, several times more effective than the political advertising campaigns spend billions on. In Canada and Poland the effect was five to ten points, and a third of it was still there 40 days later. The twist? The chatbots weren't pushy or manipulative. They won people over with facts and evidence, politely delivered. Take the facts away, and the persuasive effect roughly halves. If you work in communications, this episode is about the stakes. The same mechanics that let a state poison what AI says about a war are the mechanics that decide what it says about your brand. Understand one, and you understand the other. Guests:
Beneath the Surface is a podcast from Hard Numbers, hosted by Paul Stollery. New episodes monthly. (Ish.) Beneath the Surface is a podcast from Hard Numbers, hosted by Paul Stollery. New episodes monthly. | |||
| Is ChatGPT lying about you? | Beneath the Surface Ep.1 | 21 Apr 2026 | 00:37:30 | |
Why do LLMs 'lie'? And what can you do about it? ChatGPT told the world that Brian Hood was a convicted criminal. He wasn't. He was the whistleblower. Brian Hood spent years trying to expose a bribery scandal at a subsidiary of Australia's Reserve Bank. He raised it with the board. He raised it with the deputy governor. He was made redundant and escorted to the front gate. When prosecutions finally came, eleven people were found guilty. Brian was the key witness. Then, in 2023, ChatGPT wrote his story – and got the ending completely wrong. It said he'd been charged, found guilty, and sentenced to 30 months in jail. Brian nearly became the first person in the world to sue OpenAI for defamation. In this episode, he tells us what happened next. But Brian's story is really a window into a much bigger question: how do large language models actually decide what to say? These platforms don't retrieve facts the way a search engine does. They predict what sounds right – and when the pattern fits but the fact doesn't, you get confident fiction. We break down the mechanics in plain language: what tokens are, how pattern matching works, why a system trained on the entire internet can still fabricate a criminal conviction, and what the term "hallucination" actually gets wrong. Then we go deeper. If these systems can get things wrong by accident, what happens when people try to make them get things wrong on purpose? We speak to a senior figure from GCHQ's National Cyber Security Centre about the state-backed groups and organised criminals actively trying to poison what AI platforms tell you. He explains how little data it takes to contaminate a model's training, why the long tail problem from SEO is even more dangerous in the age of generative AI, and what the emerging discipline of defence actually looks like. If you work in communications, reputation, or GEO – generative engine optimisation – this is the foundation. You can't influence what these platforms say about your brand until you understand how they work, where they break, and who else is trying to shape the answers. #GEO #GenerativeEngineOptimisation #ChatGPT #AISearch #Reputation #LLM #Hallucination #Communications #MarComms Beneath the Surface is a podcast from Hard Numbers, hosted by Paul Stollery. New episodes monthly. | |||