Categories compared
The same counts for every category, side by side, on one ChatGPT surface at a time.
| Free | Autos → | Beauty & fitness → | Business → | Finance → | Home & garden → |
|---|---|---|---|---|---|
| How much it searches | |||||
| 90% 18 of 20 | 100% 20 of 20 | 80% 16 of 20 | 100% 20 of 20 | 100% 20 of 20 | |
| 1.0 19 in all | 1.0 20 in all | 0.8 16 in all | 1.2 23 in all | 1.0 20 in all | |
| 1.9 1.1 without the self-report | 2.8 1.3 without the self-report | 1.6 1.4 without the self-report | 2.4 1.9 without the self-report | 2.7 1.0 without the self-report | |
| 28% 5 of 18 | 20% 4 of 20 | 63% 10 of 16 | 40% 8 of 20 | 16% 3 of 19 | |
| Search journeys | |||||
| 20% 4 of 20 | 20% 4 of 20 | 50% 10 of 20 | 40% 8 of 20 | 15% 3 of 20 | |
| 5% 1 of 20 | 0% 0 of 20 | 0% 0 of 20 | 0% 0 of 20 | 0% 0 of 20 | |
| 0% 0 of 20 | 0% 0 of 20 | 0% 0 of 20 | 15% 3 of 20 0% without the self-report | 0% 0 of 20 | |
| 65% 13 of 20 | 80% 16 of 20 | 30% 6 of 20 | 45% 9 of 20 60% without the self-report | 80% 16 of 20 | |
| 0% 0 of 20 | 0% 0 of 20 | 0% 0 of 20 | 0% 0 of 20 | 5% 1 of 20 | |
| 10% 2 of 20 | 0% 0 of 20 | 20% 4 of 20 | 0% 0 of 20 | 0% 0 of 20 | |
| Query kinds | |||||
| 37 queries | 55 queries | 31 queries | 48 queries | 54 queries | |
| 3% 1 of 37 0% without the self-report | 0% 0 of 55 0% without the self-report | 0% 0 of 31 0% without the self-report | 19% 9 of 48 0% without the self-report | 0% 0 of 54 0% without the self-report | |
| What the web can move | |||||
| 93% 127 of 136 | 84% 64 of 76 | 69% 77 of 111 | 81% 93 of 115 | 87% 75 of 86 | |
| 70% 14 of 20 | 80% 16 of 20 45% without the self-report | 50% 10 of 20 | 75% 15 of 20 | 90% 18 of 20 80% without the self-report | |
| 74% 39 of 53 | 49% 33 of 67 | 73% 36 of 49 | 66% 62 of 94 | 82% 70 of 85 | |
| 87% 116 of 134 93% without the self-report | 81% 59 of 73 86% without the self-report | 69% 77 of 111 | 73% 82 of 112 83% without the self-report | 83% 72 of 87 87% without the self-report | |
| 0% 0 of 53 | 6% 4 of 67 | 33% 16 of 49 | 18% 17 of 94 | 4% 3 of 85 | |
| 16% 3 of 19 | 15% 3 of 20 | 19% 3 of 16 | 4% 1 of 23 | 10% 2 of 20 | |
| What reaches the answer | |||||
| 2.7 53 in all | 3.4 67 in all | 2.5 49 in all | 4.7 94 in all | 4.3 85 in all | |
| 0% 0 of 255 | 0% 0 of 316 | 0% 0 of 295 | 0% 0 of 388 | 0% 0 of 251 | |
| 0% 0 of 53 | 0% 0 of 67 | 0% 0 of 49 | 0% 0 of 94 | 0% 0 of 85 | |
| 55% 11 of 20 | 60% 12 of 20 | 95% 19 of 20 | 85% 17 of 20 | 85% 17 of 20 | |
| 2.8 56 in all | 3.2 64 in all | 6.3 125 in all | 3.6 72 in all | 4.0 80 in all | |
| 98% 55 of 56 35 homepages | 98% 63 of 64 17 homepages | 95% 119 of 125 89 homepages | 75% 54 of 72 30 homepages | 94% 75 of 80 24 homepages | |
| 7.2 143 in all | 3.8 76 in all | 6.6 131 in all | 5.8 115 in all | 4.5 89 in all | |
| 70% 14 of 20 | 40% 8 of 20 | 40% 8 of 20 | 30% 6 of 20 | 45% 9 of 20 | |
Each column is the category page's “At a glance” for this surface; the category's prompt cards add up to it. Every prompt is a commercial search typed word for word as US searchers type it, so the numbers hold for prompts like these, not for ChatGPT in general. One run per prompt and surface, so small differences between categories can be chance. Query, journey and query-based web numbers use ChatGPT's own list of its searches (the self-report asked for at the end of every prompt, because the app hides the second-layer fan-outs it runs while reasoning) where it contains every query seen in the stream; the number without it is shown below where it differs.