Your shopping cart is empty!
Companies already repositioned, but ChatGPT still gives old answers
Publication date: 29.09.2026
In short
- Semrush described a pattern it calls "LLM positioning lag" — when a company has already repositioned, but language models still answer with outdated information.
- ChatGPT still describes Hotjar as a standalone heatmapping tool, even though Contentsquare acquired and merged it back in July 2025.
- Capgemini found that the share of consumers who trust generative AI recommendations when choosing a product grew from 25% in 2023 to 58% in 2025.
Semrush's blog described a problem it calls "LLM positioning lag" — the gap between what a company actually does now and what language models still say about it. A business changes direction, and ChatGPT, Gemini or Perplexity keep repeating the old version for months or years.
Hotjar merged with Contentsquare months ago, and ChatGPT hasn't noticed
The clearest example in the piece is Hotjar, the behavior-analytics tool that Contentsquare acquired and folded into its own platform back in July 2025. Press releases, blog posts, YouTube videos and a dedicated landing page, live since 2021, all cover the merger. ChatGPT still describes Hotjar as an independent heatmapping tool and never mentions Contentsquare.
The same pattern shows up with Blue Apron, the meal-kit service that added pay-as-you-go ordering alongside its subscription in August 2025 and announced it on Good Morning America. ChatGPT still talks only about the "flexible subscription." Duolingo added math in 2022, music in 2023 and chess in 2025, yet ChatGPT recommends it only for language learning. Consultant Peep Laja left CRO work behind around 2016 to focus on B2B strategy at his own company, Wynter, but chatbots still name him first when asked who to learn CRO from.
Where the 58% figure comes from, and what it leaves out
Semrush cites Capgemini research showing the share of consumers who rely on generative AI advice when choosing a product or service grew from 25% in 2023 to 58% in 2025. Capgemini doesn't specify sample size or which countries were surveyed, so how well that number fits the Ukrainian market is hard to say.
The logic is simple: the more people ask a chatbot for advice, the more an outdated answer costs a business. Semrush recommends updating not just the website but directory listings, LinkedIn and Google Business Profile, and tracking what models actually say about a brand, through something like its own Perception report. We wrote about a similar approach in a guide on measuring brand visibility in AI search, and it shows that finding a mention doesn't mean a model will update what it says right away.
What struck me most is Peep Laja's case: he publicly announced years ago that he'd moved on from CRO, and chatbots still call him the go-to expert anyway. I suspect this kind of lag is even harder to fix for personal brands than for companies — there's no PR team quietly updating press releases in the background.
