The short answer
AI assistants forget brands with inconsistent messaging because a language model retrieves what is repeated and coherent across sources, not what is published most recently or most often. When a brand's website, founder, sales deck and ads each describe it differently, the assistant has no single description to reproduce, so it reaches for a competitor whose story is easier to state.
How does an assistant decide what to say about a brand?
Two ways, and both reward consistency. Without web search, the assistant answers from training recall: what it absorbed about the brand from everything public. With web search on, it reads a handful of pages and summarizes them. In both cases the assistant is looking for a description it can state confidently. A brand that is described the same way in ten places gets stated. A brand described ten different ways gets a hedge, or gets left out.
This is why the answer to "best tools for X" often includes a brand that is smaller than yours. Its story was easier to pin down.
What does drift look like in practice?
A-square's Positioning analysis reads a brand's website, 90 days of social posts, bios, ads, review-site profile and intake answers, then groups every claim into themes and gives each a share. A typical first run finds three to six competing themes, with the dominant one carrying well under half of the content. The founder's answer, the marketing team's answer and the sales team's answer to "what do we do" usually agree on a core and drift on everything else.
Each drift is a reasonable decision in isolation: a campaign for a new feature, a conference talk on a broader theme, a landing page written for a different segment. Together they make the brand unstatable.
Why does this matter more now than it did?
Because the buyer's first question is now asked of an assistant. 51% of B2B software buyers start research with an AI chatbot (G2, 2026), and 69% chose a different vendor than they first intended based on chatbot guidance, with a third buying from a brand they had never heard of (G2, 2026). Being unstatable used to cost efficiency. It now costs the shortlist.
What can a brand do about it?
Three things, in order.
- Find the one message. Not the tagline. The one claim the brand can make that competitors cannot, supported by what customers actually say they need. That claim is your brand narrative.
- Make everything carry it. Website, bios, deck, ads, the founder's next post. Not identical words, one identical claim.
- Measure whether it is being remembered. Ask the assistants the category questions a buyer would ask, with no brand name in them, and record whether the brand is named, how high, and for the right reason. Do it every week, on the same questions, with and without web search.
The third step is what A-square's Memory Score does, as one of four signals. The Memory Score page has the full definition and a glossary of the terms used here.
