Two axes, four boxes, a scatter of vendor names. If you have spent any time in B2B software, you have read a chart like this a hundred times. That familiarity is useful, and it is also a trap. An AI-generated market quadrant looks like an analyst quadrant, but it measures something different, moves for different reasons, and answers different questions. Read it the old way and you will draw the wrong conclusions.
This is a practical guide to reading a QuadrantX category quadrant: what each axis actually measures, how the boxes are defined, why vendors move, what to look at beyond the dots, and the questions the chart is not designed to answer.
Step 1: Read the Axes as Presence and Perception
The horizontal axis is Narrative Dominance, a score from 0 to 100. It captures how prominently AI models mention a vendor when asked about a category: how often the vendor appears, how early in the answer, and how strongly it is recommended. A vendor at 100 is the first name every model reaches for. A vendor at 20 is mentioned occasionally, late, and in passing.
The vertical axis is Sentiment, also 0 to 100. It captures how favourably the models describe the vendor: the balance of strengths and weaknesses they cite, the explicit language they use, and the confidence with which they recommend it.
AI Market Quadrant — A chart that positions every vendor in a category by two measures derived from AI model responses: Narrative Dominance (how prominently AI mentions the vendor) on the horizontal axis and Sentiment (how favourably AI describes it) on the vertical. It maps what buyers will be told, not what experts have concluded.
Neither axis measures product quality directly. A vendor can have an excellent product and a low Narrative Dominance because little has been written about it; another can have a middling product and a high score because it has been written about constantly. The chart is a measurement of the AI recommendation layer, which is exactly why it is useful: that layer is what a growing share of buyers consult first.
Step 2: Locate the Thresholds
The four quadrants are divided at a score of 60 on each axis. Leaders score 60 or higher on both. Challengers have Narrative Dominance of 60 or higher but Sentiment below 60: highly visible, but with reservations attached. Niche Players are the reverse: well regarded when mentioned, but not mentioned often. Laggards are below 60 on both.
The thresholds are fixed, which has a consequence worth understanding. A vendor sitting at a Narrative Dominance of 61 and one at 59 are in different quadrants and almost identical positions. When you see a vendor change quadrant between editions, look at how far it moved, not just which box it landed in. A two-point shift across a threshold is a rounding event. A twenty-point shift is a story.
Scores are also normalised so that the chart is readable: Narrative Dominance is spread across roughly 15 to 100 and Sentiment across roughly 25 to 95, which keeps the laggards visible and stops the leaders from stacking in the corner. Because of this, a vendor's absolute score is less informative than its position relative to the other vendors in the same category.
Step 3: Check How Many Models Mention Each Vendor
Behind every dot is a count that does not appear on the chart itself: how many of the five production models mentioned the vendor at all. This is the single most important number for judging how much to trust a position.
Narrative Dominance is weighted by model coverage, so a vendor mentioned by all five models earns a higher score than one mentioned equally prominently by one. But the weighting does not tell the whole story. Across all the categories we track, 41% of vendors are mentioned by only one model. A vendor named by one model with a high score is a signal that one system has a strong view. A vendor named by five is a settled fact about the market. Treat the first as a lead to investigate and the second as a position to plan around.
A dot on the chart is a summary. The number of models behind it is the confidence interval.
Step 4: Compare the Per-Model Views
The consensus quadrant is built from five separate quadrants, one per model, and the differences between them are often more informative than the average. In roughly four categories in ten, all five models agree on the leader. In the rest, they do not, and where they disagree, the reason matters.
- All models agree. The category narrative is settled. Positions will be stable, and moving them takes sustained effort over many months.
- The retrieval model disagrees. Perplexity's Sonar Pro searches the live web before answering. When it names a vendor the other four do not, recent content is being rewarded, and the other models tend to follow later if the narrative holds.
- The models split on what the category is. In some categories, different models interpret the label differently and recommend from different competitive sets. A vendor missing from two models' views may simply be in a category those models define another way.
- One model diverges on one vendor. Usually an entity or source problem: that model has read a description of the vendor the others have not. Worth checking where that model's picture comes from.
The dashboard shows the per-model breakdown for every vendor. When a position surprises you, that breakdown is the first place to look. For the full argument on why, see why multi-model consensus outperforms single-source analysis.
Step 5: Read the Edition History
QuadrantX regenerates each category on a rolling schedule, so every quadrant has a history. That history separates the categories where a position means something from the categories where it is still forming.
In roughly half of our categories, the same vendor has led in more than nine editions out of ten this year. A leader like that is not going to be displaced by a single quarter's marketing. In the other half, the lead changes hands regularly, sometimes in most consecutive editions. Those are categories where the AI narrative is unsettled, and where a vendor with a clear, consistent, specific story can move quickly.
Two habits follow. When you see a leader, check how long it has been the leader. When you see a mover, check whether it has moved before, and back. Single-edition movement in a volatile category is noise. Sustained movement in a stable category is news.
Step 6: Ask What the Chart Cannot Tell You
An AI market quadrant is not a product evaluation, and it is important to be precise about what it leaves out.
- It does not measure capability. The models have not used the products. They have read about them.
- It does not measure customer satisfaction directly. Sentiment reflects how AI describes a vendor, which tracks public reputation, including review sites, but is not a survey of customers.
- It does not know about pricing, contracts, or roadmaps beyond what has been published, and published pricing is the fact AI most often gets wrong.
- It inherits the biases of its sources. Well-documented vendors are over-represented; vendors with quiet, satisfied enterprise customers and little public footprint are under-represented. This is a property of AI recommendations in general, and it is precisely the bias a vendor needs to know it is subject to.
Read a quadrant in this order: axes, thresholds, model count, per-model split, history, and only then the headline. A vendor's dot tells you where AI places it today. The five numbers behind the dot tell you whether to believe it, and the edition history tells you whether it will still be true next month. Use the chart to understand what buyers are being told, and use direct evaluation to decide whether they are being told the truth.
Reading Your Own Position
If you are a vendor looking at your own category, three questions cover most of what matters. Are you mentioned by all five models, or by some? Is your Sentiment in line with the leaders', or is it the axis holding you back? And has your position been stable across editions, or does it swing? The answers point to three different kinds of work: presence, perception, and consistency. The chart cannot do the work for you, but it can tell you which kind to do first. You can view your category on the QuadrantX Explore page, or read how the quadrant model itself has changed in the age of AI.