AI assistants leave us out, or get us wrong
In one line
when AI assistants answer buyers' questions without mentioning you, or describe you wrongly, part of the problem corrects in weeks, part waits on timetables nobody controls, and an honest proposal says which is which.
You've probably run the test already. You asked an AI assistant the question a buyer would ask when they have the problem you solve, and back came a shortlist without you on it. Or you were there, described as smaller than you are, out of date, or filed in the wrong category altogether. The answer arrived in a confident paragraph either way; confidence is the one thing these systems never run short of. A machine just gave your first sales meeting without you, and it didn't go well.
Buyers now do this routinely. When G2 surveyed 1,076 software buyers in March 2026, 51 per cent said they begin their research in an AI chatbot more often than in a search engine. A year earlier the figure was 29 per cent.
6sense's 2025 study of more than 4,000 business buyers found 94 per cent using these tools somewhere in the purchase, mostly in the middle of it, to compare and summarise. Both are studies of software buying. If software is what you sell, they describe your buyers almost directly; if you sell services, hold the exact percentages loosely. Two separate studies pointing the same way is hard to dismiss.
Where AI assistants get their answers
The systems keep changing; the method for being in their answers holds still. An AI assistant builds its answer about your market from what it can read online. Your own site is part of that. What weighs more than most companies expect is everything published about you elsewhere: directories, reviews, press, other people's pages. If you're missing or misdescribed, what's published about you is thin, stale, or contradicts itself from one source to the next.
Sometimes it's worse than thin: a technical setting on your own site, put in place for some other reason, is turning their readers away at the door. That fault is common, invisible, and gets checked first, because corrections published behind a closed door reach nobody.
The thinness is in what's published about your work where machines now read, and says nothing about its quality.
The three ways an answer goes wrong
Plainly wrong facts
the wrong city, the wrong size, a service you don't sell. These usually trace back to one stale source, and they're the fastest to fix.
Right facts, undersold
everything stated is true, and the impression is of a company smaller or narrower than yours. Be careful buying help with this one. The impression is a summary of what the wider web says about you, so it can't be corrected on your own pages alone. Any proposal claiming your website by itself will fix how AI assistants describe you is wrong, from anyone.
The wrong box
the answer files you in a category you're not in, and it gets corrected in the directories and listings the classification comes from, which sit outside your site too.
The two speeds of correction
Some answers are looked up fresh: the AI assistant fetches sources at the moment of asking and writes from them. Fix the sources and these answers move in weeks. Most of the correctable work lives here.
Other answers come from the model itself, from what it learned when it was trained. Those change when the AI assistant's maker retrains its model after the sources were corrected, and nobody outside that company controls when that happens, Brandian included. An honest proposal states, for every fault it finds, at what speed it corrects, so you're never promised movement on a schedule your supplier doesn't set.
The work Brandian refuses
Some of what's sold as AI visibility work is written for machines: pages stuffed with self-description for systems to quote, league tables planted around the web in which the author finishes first, mentions bought to look earned. Brandian refuses all of it and writes one version of your company, for the person reading it, with no second version prepared for the machines. What's left is slower and checkable: fixing what's published about you, and measuring whether the answers moved.
One test you can run this week
You can take your own baseline today, without anyone's help. Borrow a tester: someone who has never mentioned your technology company to their AI assistant. Have them ask it what your company does and what it sells, and the same for the two competitors you most often lose to. Save the answers with the date on them.
Where to start
Your saved answers give you a baseline. What they can't tell you is why each answer went wrong, which faults will move, or how soon. Brandian has two ways to find out, and the findings are yours whichever you choose.
The AI Visibility Inspection examines how five AI assistants answer the questions your buyers ask. It states what's wrong in each answer, at what speed each fault corrects, and whether anything on your own site is blocking the fix. The corrections themselves are a separate piece of work: the Inspection ranks them and puts them in writing, and who makes them is your decision.
The Brand Record is the other door in: the measure of how your buyer sees your brand today, AI assistants' answers included, which everything that follows is checked against. Whichever you buy first, its full price comes off the first instalment of the service that includes it, inside six months.
£2,200
£4,000
Comments