Record · the firm measured on itself
We ran our own method on ourselves. We were not in the answer.
Taken 30 August 2026. Singapore. ChatGPT and Gemini consumer apps, fresh sessions, memory off. Every quote below is verbatim and you can re-run any of it.
Elstrand sells one thing: we find out whether AI assistants recommend your brand when a shopper asks what to buy.
So a fair question is whether they recommend us.
They don’t.
The question
Asked exactly as a small brand owner would ask it:
I run a small consumer brand in Singapore. I want to find out whether ChatGPT and other AI assistants recommend my brand when shoppers ask what to buy. Is there a service that tracks this for me?
ChatGPT named: Otterly AI, Peec AI, Profound, AITrack.sg, Addure.
Gemini named: Otterly.AI, LLMrefs, SE Ranking / Semrush AI Toolkit, Profound.
Elstrand appears in neither. Between the two answers, nine services were named and only two — Otterly and Profound — appeared on both lists.
The part that stings
We built this firm on the argument that a single AI answer is not a measurement. ChatGPT made that argument for us, unprompted, in the same answer that left us out:
AI answers are stochastic, so a single manual ChatGPT query isn’t a reliable measurement. Good monitoring services repeatedly run a defined prompt set and look at trends rather than treating one answer as “the ranking.”
It agrees with us completely. Then it recommends five other companies.
We also ask people not to buy another dashboard. Our home page says so. Asked that exact question, in those words, ChatGPT named e-Alchemists, Addure, CANLAH AI, AI Studio and SingRank. One competitor was described like this:
a locked set of buyer queries, repeated multi-engine probes, archived raw responses, and monthly reporting with evidence, rather than relying on a single AI response.
That is our method statement, almost word for word, credited to somebody else. Another was called “a Singapore-based managed/advisory option rather than simply a dashboard”, which is our positioning, awarded elsewhere. Its closing line named two other firms as the ones to call first.
We are known. We are just not recommended.
Ask either assistant about Elstrand by name and the tone changes completely. ChatGPT calls it “a real, functioning service, not an obvious scam, but… a very young, small business”, and credits the published methodology, the named founder, the public pricing, the free check, and the explicit refusal to guarantee placement. On the method itself:
they claim to use the actual consumer interfaces rather than developer APIs, with fresh sessions and memory turned off. That’s an important methodological distinction.
It also says what is wrong with us, and we are not going to hide it:
The sample sizes are tiny.
And that what you are buying is “research/interpretation, not statistically comprehensive monitoring”. Its own verdict table answered no to established, no to would-pay-immediately, and absolutely to would-take-the-free-check.
Both criticisms are fair. We raised the question counts in response to the first. The second is answered by time and by clients willing to be named, not by anything we can write on this page.
This is the whole product, demonstrated on ourselves
Three things happened here, and they are the three things we find for clients.
Known but not recommended. The assistant has the facts and does not reach for them unless somebody types the brand name first. We have measured this in eyewear, in lingerie, in candles, in supplements and in jewellery. It is the most common finding there is.
Positioning credited to a competitor. The words a brand uses about itself turning up in an answer, attached to somebody else’s name.
The assistants disagree with each other. On our own category, nine services named across two assistants and only two in common. We have recorded the same split in candles (1 of 4 brands in common), jewellery (2 of 4), bedding (1 of 4) and activewear (3 of 5). It has never yet gone above 60%. Anyone checking one assistant and stopping is reading a fraction of the picture.
Why we published this
Because we would rather show you the method working on an unflattering case than show you a testimonial we do not have.
If it finds this much about the firm that built it, it will find something about yours.
Every quote on this page was taken on 30 August 2026 and can be reproduced by asking the same questions in a fresh session with memory off. They will not come back identical. That is the point, and it is why we read them monthly rather than once.