The uncomfortable truth is that the machine is not choosing the better business. It cannot tell which of you is better. It is choosing whichever business it can read the most about, from the most sources, in the words the customer used.
We measure this for a living, and the firm that gets named is very often not the strongest in its market. It is the one whose evidence happens to be machine-readable. We have measured accredited specialists losing to competitors with a fraction of their experience, purely on the shape of a web page.
Four reasons they are in and you are out
- They have a page for the exact question. Where you have a services page, they have a page about the specific job in the specific town. The machine matches the question to the page almost literally.
- Their facts are quotable. Prices, turnaround times, coverage areas and qualifications written in text can be lifted straight into an answer. Yours may be in a brochure, a graphic or a conversation.
- Someone other than them says so. A trade register, a directory that gets cited in your market, a local news mention. Independent corroboration is the strongest signal there is, and it is the one most businesses have least of.
- They have been readable for longer. Some of it is simply time in the index, which is frustrating and also fixable only by starting.
The part nobody tells you: it moves
These positions are not tenured. In one market we measured across two cycles this summer, a competitor's share of the citations an assistant used fell by more than half within a month, after they rebuilt their website and every page lost the signals that told the machine which pages had changed. Nothing about their business changed. Their evidence did.
The same works in reverse, which is the encouraging half. Answers vary between identical runs about 19% of the time, and shift further whenever the platforms change how they write answers. Being absent today is a position, not a verdict.
What actually closes the gap
- Find out which questions you lose and who wins them. Not a general score, the actual questions.
- Build the page that answers the specific question, in specific terms, with the facts in text.
- Get corroborated somewhere the machines in your trade already read. Which places those are is measurable rather than guessable.
- Measure again, repeatedly, so you can tell a real change from ordinary noise.
That sequence is what our method does, and the first step of it, finding out which questions you lose and who takes them, is free before we work together.
Common questions
Why does AI recommend my competitor instead of me?
Almost always because more readable, independent evidence exists about them at the moment the answer is built. Not because they are better. The machine cannot assess quality of service; it assesses what it can retrieve and how well independent sources agree. A competitor with clear prices in text, a specific page for the question asked and a few third-party mentions will beat a better business whose evidence is thin.
Does being recommended mean they are paying for it?
No. There is no paid placement in the organic part of these answers. Some assistants are beginning to carry advertising, but it is labelled and separate from the recommendation itself.
Can I get them removed?
No, and you should be wary of anyone who implies otherwise. The only route is to make your own evidence strong enough that you are included, which is a different and more durable job.
They have fewer reviews than me. How is that possible?
Because review counts sitting on a platform's own site are weaker evidence than most owners assume. In one UK service market we measured in July 2026, review platforms accounted for 56 of 4,649 sources cited across AI answers, about 1%. What the machines mostly read is ordinary web pages.
Updated August 2026