<b>Difficulty scores disagree by 22 points on average</b>

<b>Difficulty scores disagree by 22 points on average</b>
Difficulty (KD) is a model output, not a fact. We scored the same 600 commercial queries across three vendors and measured the spread.
—
Agreement within 10 pts ▇▇▇ 28% of queries
Spread of 11–25 pts ▇▇▇▇▇ 49%
Spread over 25 pts ▇▇ 23%
—
The biggest disagreements clustered on terms with mixed SERPs (forums plus brand pages), where link-based and SERP-based models read the page very differently. Single-source KD is fine for sorting, dangerous for go/no-go calls.
<b>So what:</b> for any keyword you'd build a page around, pull KD from two models and treat the gap as your uncertainty band.
Benchmark of the week: KD spread widens to 31 points on queries where the top 5 mixes UGC and editorial.