The index
Risk is near-universal. Its concentration is not.
When AI reviews a business contract it almost always finds something. In this corpus 97.7% of contracts carried at least one issue, and 87.0% carried at least one red-flagged risk. That is close to universal, and on its own it says very little. Contracts are drafted by people under time pressure, and a review exists precisely to catch what they missed.
The useful question is not whether a contract carries risk. It is how much, and where it concentrates. Measured per hundred paragraphs of contract text, the answer varies by a factor of 1.69 between the highest industry and the lowest.
Real estate contracts carry the highest density of flagged risk of any industry measured, at 2.67 red-flagged risks per hundred paragraphs, against an all-industry average of 1.67.
GenieAI Contract Risk Density Index 2026. n = 2,320 contracts, 1,245 organisations, February to July 2025.
| Industry | Risks per 100 paragraphs | Contracts | Organisations | Tier |
|---|---|---|---|---|
| Real Estate | 2.67 | 64 | 58 | Headline |
| Media | 2.29 | 60 | 47 | Headline |
| Consumer, Public & Health Services | 2.22 | 77 | 58 | Headline |
| Sport & Entertainment | 2.10 | 32 | 27 | Supporting |
| Consultancy | 1.95 | 189 | 151 | Headline |
| Energy | 1.95 | 41 | 22 | Supporting |
| Manufacturing | 1.89 | 34 | 28 | Supporting |
| Finance | 1.83 | 71 | 56 | Headline |
| Legal Services | 1.83 | 195 | 160 | Headline |
| Agriculture, Forestry & Fishing | 1.63 | 45 | 40 | Supporting |
| Education | 1.58 | 75 | 59 | Headline |
| All industries | 1.67 | 2,320 | 1,245 | Baseline |
Industries with at least 50 contracts are shown as headline; those with 30 to 49 are shown as supporting. No single organisation exceeds a third of any published industry.
Real estate and media sit at the top, and both are industries whose contracts turn on things that are hard to pin down in writing: what condition a property is in, what a licence actually covers, who owns a derivative work. Education sits at the bottom, and its contracts are the most standardised in the set.
Why we measure per page
Counting problems rewards short documents.
The obvious way to rank industries is to count problems per contract. It gives the wrong answer, and legal services shows why.
Contracts belonging to legal services organisations have the highest ninetieth-percentile risk count of any industry measured, at 13 red flags, and are among the least risk-dense once contract length is taken into account.
GenieAI Contract Risk Density Index 2026. n = 195 contracts, 160 organisations.
On a raw count legal services would rank near the top. Measured per hundred paragraphs it sits joint-bottom at 1.83. The difference is length. Counting raw problems rewards short documents and punishes thorough ones, which is exactly backwards: a long contract that spells out its terms carefully is doing the thing we want. Density asks the question a buyer actually has, which is not how many problems are in this document but how much of this document should worry me.
It also separates two measures that are usually assumed to move together, and mostly do not.
| Industry | Risk density | Density rank | Flag rate | Flag-rate rank |
|---|---|---|---|---|
| Real Estate | 2.67 | 1 | 96.9% | 2 |
| Media | 2.29 | 2 | 98.3% | 1 |
| Consumer, Public & Health Services | 2.22 | 3 | 93.5% | 3 |
| Consultancy | 1.95 | 4 | 89.4% | 6 |
| Finance | 1.83 | 5 | 91.5% | 4 |
| Legal Services | 1.83 | 5 | 91.3% | 5 |
| Education | 1.58 | 7 | 84.0% | 7 |
Headline-tier industries only. Finance and legal services share a position at 1.83 density and a flag rate of roughly 91%.
Media contracts are the most likely to contain a problem. Real estate contracts contain the most problems per page. Only education is bottom of both. Flag rate tells you how likely it is that something was missed; density tells you how concentrated the misses are once you find them. This index ranks on density and reports flag rate alongside it.
What the index leaves out
The excluded industries account for more contracts than the published ones.
An index that shows only what flatters it is not worth reading, so here is what is missing and why.
The rule, set before the numbers were run
No industry is published where a single organisation accounts for more than a third of its contracts. Technology and construction both fail it, and both would have been the more interesting rows.
Technology is absent, and it is the largest industry in the corpus at 922 contracts, more than a third of everything reviewed. 70.2% of them come from a single organisation, so any figure we published would describe one company's contract portfolio and call it an industry. Construction is absent for the same reason, at 42.1% from one organisation. Six further industries have too few contracts to report at all.
| Group | Contracts | Share | Reason |
|---|---|---|---|
| Eleven published industries | 883 | 38% | Published |
| Technology | 922 | 40% | One organisation supplies 70.2% |
| Unclassified | 301 | 13% | No single declared industry |
| Construction | 126 | 5% | One organisation supplies 42.1% |
| Six smaller industries | 88 | 4% | Fewer than 30 contracts each |
Unclassified contracts belong to organisations that declared more than one industry or none.
Together the exclusions account for 1,437 of 2,320 contracts. The eleven published industries cover 883. Anyone who adds up the table will find that gap, and it is better stated here than discovered later.
What actually gets changed
Contracts are reworked narrowly, not deeply.
A separate and much larger population, and a different question: when businesses do rework a contract, how much of it do they touch? Very little.
Of contracts that were negotiated at all, 72% were marked up in exactly one clause, and the median contract had one clause reworked.
GenieAI Contract Risk Density Index 2026. n = 10,682 contracts, 9,520 organisations.
| Measure | Global | United Kingdom | Rest of world |
|---|---|---|---|
| Median clauses marked up | 1 | 1 | 1 |
| 90th percentile | 4 | 5 | 3 |
| 95th percentile | 7 | 12 | 6 |
| 99th percentile | 63 | 103 | 40 |
| Most in a single contract | 385 | 294 | 385 |
| Marked up in exactly one place | 72.1% | 62.9% | - |
| More than 60 clauses marked up | 1.06% | 1.69% | 0.80% |
Clause-level negotiation activity. UK n = 2,247 contracts; rest of world n = 6,002.
UK businesses rework contracts more broadly than the rest of the world, though the difference sits at the margin rather than the middle: 37% of UK contracts had more than one clause reworked, against 28% elsewhere, and the most heavily reworked tenth touched five clauses or more against four. At the far end sits a single contract with 385 separate clauses marked up, which is less a negotiation than a rewrite.
Two populations, not one
The contracts in this chapter are largely not the contracts in chapters one to three. The two populations overlap by only 11%: reviewing a contract is not what causes it to be negotiated. These figures describe the breadth of a negotiation, meaning how many clauses were touched, and never its depth. The data records which clauses were marked up, never how many times.
About this data
About this data
Figures are aggregated, anonymised telemetry from contracts reviewed on the GenieAI platform. No contract text, personal data or customer-identifiable information is included. A risk is a red-flagged finding; an issue is a red or amber finding. Density is red-flagged risks per hundred paragraphs of contract text. Industry is self-declared by the organisation at signup from a fixed list, and figures count only organisations that declared a single industry. Review times are flat across every industry, between 22 and 29 seconds at the median, so they are not reported per industry. Chapter four draws on a separate population of contracts with clause-level negotiation activity, which overlaps the review corpus by 11%.
| Sample | 2,320 contracts from 1,245 organisations for the index. 10,682 contracts from 9,520 organisations for the negotiation chapter. GenieAI has handled 381,109 contracts from 247,303 organisations since June 2020, of which 2.6% have ever had a risk review. |
|---|---|
| Period | 19 February to 31 July 2025. |
| Data as at | September 2026 |
- No industry is reported below 50 contracts, or fewer than 5 organisations, or where a single organisation accounts for more than a third of the contracts. Four industries are reported at a lower threshold of 30 contracts and are marked as supporting.
- Review records before 19 February 2025 were removed under data retention and no longer exist.
- From August 2025, findings stopped being linked to a specific paragraph of the source document, so risks per hundred paragraphs cannot be computed after that date. This index is a fixed study of a closed period rather than a running tracker.
- Only 2.6% of contracts on the platform have ever had a risk review, so this is a selected population and not a sample of all contracts.
- All figures were read from a single database snapshot on 3 September 2026.
Cite this report
Use these figures
You’re welcome to reproduce these statistics and charts with attribution and a link to this page.
Cite as
GenieAI (2026). Contract Risk Density Index. https://www.genieai.co/research/contract-risk-density-index-2026 