Dan Taylor of 10 Spurs wrote a column for Search Engine Journal, published on 12 November 2026. His argument is that AI visibility tools should not lead your AI search strategy. He says the market is young and unstable, and that dashboards can give leaders false comfort. He counts 194 AI visibility tools launched in the past two years. About 60% of them are lightweight experiments, side projects or solo builds, and he estimates 55 to 60% of these startups will fail. His advice is to own your prompts, keep your raw data and measure visibility yourself.
The excitement is about the dashboard
Every new channel gets a wave of tools that promise to measure it. AI search is no different. The pitch is easy to like. Pay a monthly fee, see a score, watch the line go up.
The question I ask first is whether there is a real need. There is. A business does need to know whether ChatGPT, Gemini or Google’s AI Overview names it when a customer asks a buying question. That need is real and it is new.
What the tool adds on top is less clear. Taylor points out that tools measure visibility differently from each other. One tool may report a brand as number one in 80% of answers while another misses it entirely. Both can be running correctly and still disagree. A score from a tool is only meaningful inside that tool’s own method, and that method can change, or the vendor can disappear.
Nothing about the underlying problem changed
Here is what is true whichever tool you use. AI answers are regenerated on every search and they change week to week. Taylor makes the same point. Much of the movement in citations between runs is statistical noise, so a team that reacts to every wobble wastes its time.
That is why I tell people to date every result. An answer with no date attached is nearly useless. “We were named in the AI Overview” means little. “We were named on 12 November, in a Singapore browser, for this exact question” is a fact you can compare against next month.
What also did not change is how brands get named. The AI picks the brand it understands most clearly and can quote most precisely. No dashboard creates that. Work on your pages and on the other sources that describe you does.
What a small business should hold itself
Picture a clinic, a shop or a service business with one person looking after marketing. The tool they pick this year might be gone or repriced next year. If their only history lives inside that tool, they lose it.
So they should keep three things in their own files:
- The prompts. A list of the questions a real customer would ask when looking for what you sell. Include comparison questions and a few that name your brand. Taylor suggests several types of prompt set, including customer-intent, comparative, category-defining and brand. You decide the list. A vendor should not decide it for you.
- The raw answers, with dates. Save the full text of each answer, the date you asked, the engine and the sources it listed. Taylor recommends keeping response text, timestamps, source URLs and prompt inputs in your own infrastructure so the history survives a change of vendor.
- A simple record of whether you were named. One row per prompt per date. Named, not named, and who was named instead. Being named in the answer is the commercial result. Being listed as a source is a useful early sign, but it is the second thing to record.
A spreadsheet is enough for this. If a tool can export raw data cleanly, use it as an extra counter. Taylor says to favour vendors that offer raw exports, and I agree. If a tool cannot export, treat that as a reason not to depend on it.
Let the machine count and keep the judgement
Running the same twenty prompts every week and logging the results is tedious. That is a good job for a machine. Counting how often you were named, checking the sources, drafting the weekly summary are all things I let Claude do.
The judgement stays with the person who knows the business. Which questions matter to your customers? Is a drop in naming real or just noise across repeated runs? Is the fix on your own pages, or is nobody else on the internet describing you? A tool cannot answer those for you, and a score on a dashboard does not either.
This is also why I put dates on my own evidence. Alva’s own coffee store was listed first on 3 of 5 commercial queries on 30 and 31 August 2026, in a Singapore browser. Without the dates and the scope, that line would say far less. It is a snapshot, and I would not want anyone to read it as a permanent position.
What to do this month
- Write down 15 to 25 questions your customers would ask an AI when choosing a business like yours.
- Run them in the engines your customers use. Save the full answers with the date.
- Record, for each one, whether you were named and who was named instead.
- Repeat every week or two. Look at the trend over several runs and ignore any single result.
- Use a tool only if it exports raw data, and keep your own copy anyway. Don’t plan next quarter around a feature a vendor has promised.
If a tool shuts down tomorrow, you should lose a counter and nothing else.
Source: Search Engine Journal: Don’t Let A Tool Lead Your AI Search Strategy – Own Your Data Instead