You just paid for a repositioning, updated the homepage and briefed the sales team. Then a buyer asks an AI assistant for the best option in your category and the answer describes the business you ran three years ago. Semrush calls that gap the LLM positioning lag.
It matters because AI answers now sit in front of most purchase research. A Capgemini figure cited in the piece puts consumers who replaced search engines with generative AI for product and service recommendations at 25% in 2023 and 58% in 2025. If your AI profile is stale, more than half your potential buyers may be learning the wrong thing about you.
The lag shows up even when PR did everything right
Semrush tested brands and individuals whose repositioning has been public for years. Hotjar merged into Contentsquare by July 2025, yet ChatGPT still recommended Hotjar as an independent heatmapping tool more than a year later, with no mention of Contentsquare. Blue Apron announced a pay-as-you-go option in August 2025 through Good Morning America and national press, but ChatGPT still described a subscription-only service. Duolingo has taught math since 2022, music since 2023 and chess since 2025, but the assistant still framed it only as a language app. And Peep Laja, who moved from CRO to B2B strategy around 2016, still appeared first in a ChatGPT list of CRO experts.
The pattern is clear: years of old positioning can outvote a year of new messaging in a model’s training data.
What keeps the model stuck
- Training cutoffs: a change that happened after the model last trained may not exist at all unless the model decides to search.
- Conflicting signals: your current site, old directories and stale profiles describe different brands, and the model has to guess which one is right.
- Volume effect: ten years of articles, reviews and backlinks describing the old position outweigh a new press release.
How to make the AI answer match the new you
The fix is not one PR blast. It is coordinated work across PR, content and brand monitoring.
First, publish a dated, explicit announcement on your own domain and distribute it. Say what you do now, who you serve, and what changed. Then audit old pages and authoritative inbound links, and request updates where your old category or product name still appears. Keep homepage, about pages, meta descriptions and internal anchor text consistent, and keep Google Business Profile, LinkedIn, G2, Crunchbase, Wikipedia and directories current. Structure company pages so each section can stand alone, with headings that match how buyers actually ask. Finally, monitor what LLMs say about you over time, because missing a mention is as costly as an inaccurate one.
If you have a monitoring tool, use it. If not, ask the questions a buyer would ask and track the answers monthly. Think of it as brand hygiene for the AI layer: your website can be perfect, but if the model still quotes the old version, the buyer never arrives.
Source: Semrush Blog
