The Method
The Recommendation
Baseline
How do I get AI models to recommend me?
You get recommended when an AI system can identify who you are, verify what you do, and find that description confirmed somewhere it already trusts. Recommendation follows verification. The Recommendation Baseline is how Start Solutions AI measures whether that is happening today: a fixed set of the questions your buyers actually ask, run across ChatGPT, Perplexity, Gemini, and Google AI Overviews, recorded verbatim, and scored the same way every month.
What a baseline is
A baseline is a measurement you can repeat.
Most AI visibility work stops before this point. One screenshot of ChatGPT naming a competitor is an anecdote. It cannot tell you whether last month was better, whether a change worked, or whether the engine simply answered differently that day.
The Recommendation Baseline fixes three variables so comparison is possible.
The questions
Eight to fifteen questions your future clients ask before they choose someone. Not keywords. Not the questions you worry about.
The engines
ChatGPT, Perplexity, Gemini, and Google AI Overviews. The same four, every run.
The scoring
Four measures, applied the same way each month.
Answers are recorded verbatim. The exact wording is the evidence. Summarizing it would destroy the thing being measured.
The question set
There are two kinds of questions and only one of them belongs in a baseline.
Anxiety questions are the ones you ask. Do I still need SEO. Is my social media enough. Am I behind. Those are worth answering in your content. They are not what your buyer types.
Buying questions are what your future client asks in the minutes before they choose someone. They take three shapes.
Category
Best [category] for [type of buyer].
Problem
The specific problem you solve, phrased the way someone living with it would phrase it.
Verification
Your name, your reviews, whether you are legitimate.
Every set is built for one business, from that business's city, specialty, and buyer type. Question sets are not shared between clients, and one client's results are never used to describe another.
What gets scored
Recommendation rate
How often you are named across the full question set. This is the headline number and usually the one that surprises people.
Who gets named instead
The names the engines return when yours does not appear.
What the answer cites
The sources the engine drew from to build its answer. This is where the work goes.
How you are described
When you are named, whether the description matches what you actually do.
These are separate measurements and they routinely disagree. One leadership speaker scored 83 on description quality and appeared in zero percent of recommendation answers. The engines knew who he was. They did not put him in the answer.
A strong description with a zero recommendation rate is not a good result. It is a specific one, which makes it fixable.
How often it runs
Run once, it is a snapshot. Run monthly against the same questions and the same engines, it is a baseline.
The re-score uses the original question set. New questions can be added to a new set. The original is never edited, because editing it breaks the comparison, and the comparison is the only thing that makes the number mean anything.
What this does not claim
There is no ranking scoreboard in AI search. No position one. No placement anyone can verify for you.
- We do not promise that any engine will name you.
- We do not promise a position, because the engines do not publish one.
- We do not report a number without the verbatim answer behind it.
- We do not treat one good answer as a result.
Anyone who promises rankings in AI search cannot show you how they were measured.
Common questions
How do I get AI models to recommend me?
An AI system recommends what it can verify. It has to identify who you are, understand what you do, and find that same description confirmed in sources it already trusts. The Recommendation Baseline measures whether that is happening today, then the work builds the evidence that changes the answer.
What is The Recommendation Baseline?
It is the method Start Solutions AI uses to measure AI visibility. A fixed set of eight to fifteen buyer questions, run across ChatGPT, Perplexity, Gemini, and Google AI Overviews, recorded verbatim, and scored the same way each month.
Which AI engines are measured?
ChatGPT, Perplexity, Gemini, and Google AI Overviews. The same four every run, so results stay comparable month over month.
How many questions are in a baseline?
Between eight and fifteen. Enough to see a pattern rather than a single result, few enough that every answer can be read rather than summarized.
How often is it re-run?
Monthly, on the same question set. The first run is the baseline. Every run after that is measured against it.
Is this the same as SEO?
No. SEO optimizes pages to rank. AI visibility works one step earlier, on whether a system understands a business well enough to name it. The two overlap on technical groundwork and diverge on almost everything else.
Can you guarantee ChatGPT will recommend me?
No. No one can, and the engines publish no position to guarantee. What can be done is measure whether you are named today, strengthen the evidence a model would need to name you, and measure again.
See your own baseline
The first run is free. Request an AI Visibility Snapshot and see how often the engines name you today.
Request a Free SnapshotNot ready? Take the Visibility Scorecard. Eight questions, about 90 seconds. No PDF, no sales call.