The top tens we measured were not ten places a written page could win.
We bought the real Google results for 20 keywords to grade content tools against. Counting what came back answered a question we had not asked: of 200 top-ten slots, 146 held an ordinary text page. The remaining 54 went to something else.
A top ten in this fixture held 7.3 organic text results on average, 5 at worst and 10 at best. A written article could not compete for the other 27%. 4 of those 54 slots held a video. A video is organic, but an article cannot take that place either. The rest held something the snapshot does not name.
One position accounts for most of it
On 14 of 20 keywords, no organic text result held position one. Nothing else comes near that. The other nine positions lost between 3 and 7 keywords each, with no slope down the page: position two lost 4, position four lost 7. So the page is not filling up from the top. One slot behaves differently from the other nine, and every ranking claim you read measures itself against that slot.
| Position | Keywords with no text result |
|---|---|
| 1 | 14 |
| 2 | 4 |
| 3 | 3 |
| 4 | 7 |
| 5 | 6 |
| 6 | 3 |
| 7 | 3 |
| 8 | 5 |
| 9 | 4 |
| 10 | 5 |
Strong sites hold what is left
Among the 146 text results, the median domain rating is 85.5. Just 6 of them, 4%, came from a domain rated under 50. One domain, reddit.com, appears in the top ten of 13 of the 20 keywords.
| Domain rating | Results | Share |
|---|---|---|
| Under 30 | 1 | 1% |
| 30–49 | 5 | 3% |
| 50–69 | 23 | 16% |
| 70–89 | 53 | 36% |
| 90 and up | 64 | 44% |
Every keyword
| Keyword | Organic | Median DR | Slots taken |
|---|---|---|---|
| best ai for math | 5/10 +1 video | 62 | 1, 5, 6, 8, 10 |
| best ai for writing | 5/10 | 94 | 1, 3, 5, 7, 10 |
| ai for coding | 6/10 | 77 | 1, 2, 4, 5 |
| best ai chatbot | 6/10 +1 video | 80 | 1, 2, 6, 9 |
| best ai seo tools | 6/10 +1 video | 74.5 | 1, 8, 9, 10 |
| best ai writing tools | 6/10 +1 video | 94 | 1, 4, 8, 10 |
| ai for business automation | 7/10 | 90 | 1, 3, 4 |
| ai for customer service | 7/10 | 92 | 2, 7, 8 |
| best crm for small business | 7/10 | 92 | 1, 2, 5 |
| best help desk software | 7/10 | 91 | 1, 6, 9 |
| best laptop for programming | 7/10 | 89 | 1, 5, 9 |
| ai content detector | 8/10 | 77.5 | 4, 10 |
| ai for marketing | 8/10 | 92 | 1, 5 |
| ai for small business | 8/10 | 92 | 1, 3 |
| ai for teachers | 8/10 | 85.5 | 1, 4 |
| digital nomad visa countries | 8/10 | 61.5 | 1, 7 |
| ai presentation maker | 9/10 | 83 | 4 |
| ai voice generator | 9/10 | 86 | 8 |
| cost of living comparison | 9/10 | 84 | 4 |
| gpu comparison | 10/10 | 70.5 | none |
What this does not say
A slot we count as taken is one no organic text result held. The snapshot drops AI Overview citations, People Also Ask blocks and duplicate sitelinks without recording which of them took a given position, so we can say the place was closed, not what closed it. Anyone reading a specific format into these counts is adding something the data does not carry.
The fixture is 20 English keywords chosen to test content tools, not a sample of Google: eight informational, nine commercial, three tool-seeking, no brand terms and nothing local. A different set would give a different share.
Re-running the same 20 keywords and finding the organic count back near ten. The snapshot covers one week in August 2026. Google adds and withdraws these formats continuously, and nothing here claims the share holds. The method explains how the fixture was frozen, and the per-keyword rows sit on the corpus page.
Content optimization tools score a page against the ten results they see. The scores we have measured do not predict where a page ranks. These counts sit underneath that: 27% of the target was never a place a written page could take, whatever it scored.
ToolVerdict. “The top tens we measured were not ten places a written page could win.” https://toolverdict.ai/findings/serp-real-estate. Last updated 2026-08-18.