Book a demo Log in / Sign up

GenieAI Research · 2026

The Vagueness Problem: What AI Review Reveals About UK Business Contracts

An analysis of 365,360 contracts from 244,337 organisations, including 66,234 from 32,448 UK businesses. The most common problem AI finds in a business contract is not money or liability. It is imprecision.

Jump to the findings Download the data and charts

Written by Imad Mohammed Nazar, Swetha Meenal · Data as at July 2026 · 5 min read

1 Key findings

Five things we learned from UK contract review.

14.6%
Vague wording is the most-flagged issue, every quarter
77%
of UK contracts flagged for a risk issue (vs 57% global)
+43%
more issues when a contract is vague
~2 min
median full-contract review, 95% within five minutes
2 The vagueness problem

The most common contract problem is imprecision, not money.

When GenieAI reviews a contract, it raises specific issues for a person to check. Across 37,594 such issues over 3,692 contracts, the most common category was not payment or liability. It was vague or ambiguous wording, undefined scope and open-ended deliverables.

Vague or ambiguous wording is the most common issue in UK business contracts, at 14.6% of all issues raised, ahead of payment terms (10.5%) and liability (9.8%) - and the #1 category in every quarter measured.

GenieAI UK Contract Review Report 2026. n = 37,594 issues, 3,692 contracts.

Risk categories by share of issues raised
Bar chart: vague/scope 14.6% is the largest risk category, ahead of payment 10.5% and liability 9.8%.
Risk categoryShare of issues
Vague / scope14.6%
Payment & fees10.5%
Liability & indemnity9.8%
Termination7.8%
Dispute resolution7.4%
Regulatory / compliance6.6%
Confidentiality3.9%
IP / ownership3.5%
Non-exclusive keyword classification of AI-generated issue titles.
External context World Commerce & Contracting finds an 11% erosion of contract value from poor contracting, and that only 39% of professionals believe contracts achieve their intended goals, down 40% since 2017 (WorldCC / Deloitte, 2024-2026).

The instinct is to fight hardest over money and liability, because those feel like the dangerous clauses. The more common problem is quieter: a scope left vague, a deliverable never pinned down, an obligation written loosely enough that two sides can read it two ways. Those are the terms that look harmless on the day you sign and become the dispute eighteen months later. Precision is not pedantry. It is the cheapest insurance a business can buy.

Imad Mohammed Nazar, Report reviewer, GenieAI (draft for sign-off)
3 What's in a contract

Most contracts carry more than one problem, and UK contracts more than most.

Around six in ten contracts GenieAI reviews carry at least one risk issue, and the typical flagged contract raises 9 to 12 of them. Two patterns stand out.

GenieAI flagged at least one risk issue in 77% of UK contracts reviewed, versus 57% globally, a 20-percentage-point gap.

GenieAI UK Contract Review Report 2026.

Contracts flagged for vague wording carry 43% more risk issues overall (11.2 versus 7.8 on average) than clear ones.

GenieAI UK Contract Review Report 2026.

Clause types that attract the most AI work
Clause typeShare of clause actions
Payment & fees19.8%
Termination / expiry14.6%
Liability & indemnity10.8%
Dispute resolution9.7%
Confidentiality8.3%
Substantive-clause subset, n = 2,158 paragraphs. Non-exclusive.
Enterprise-authored contracts are flagged slightly more often than SME ones (70% vs 65%). And the flag rate has fallen over the past year (67% to 48%) while flagged contracts are examined more thoroughly (median issues 9 to 12); part of this reflects simpler reviews migrating to a different GenieAI workflow.

Almost every contract we looked at had something worth questioning, and the typical one had nine separate points. The practical lesson is to check the unglamorous clauses: payment triggers, termination rights, what happens to your data and who owns the work. That is where the value and the risk actually sit.

Swetha Meenal, Report reviewer, GenieAI (draft for sign-off)
4 Reviewed in minutes

AI review is no longer the bottleneck.

The reason a business can afford to check every clause of every contract is that the review itself is no longer the bottleneck. In the most recent quarter GenieAI reviewed a median contract in about two minutes, with 95% of reviews complete within five minutes.

GenieAI reviews a full contract, clause by clause, in a median of around two minutes, with 95% complete within five minutes.

GenieAI UK Contract Review Report 2026, most recent complete quarter.

Median AI review time, by quarter
Median AI review time rising from 20 seconds in Q1 2025 to 117 seconds in Q2 2026.
QuarterMedian AI review time
Q1 202520s
Q3 202539s
Q1 202660s
Q2 2026117s
Server-side review time; excludes user think-time.
Review time has risen over the year, from around 20 seconds in early 2025 to about two minutes now. This is not the AI slowing down: simpler reviews have moved to GenieAI's newer agent workflow, leaving the classic pipeline handling progressively larger and more complex contracts. We make no 'faster than a human' claim; the point is simply that a full clause-by-clause review now takes minutes, not the hours or days it would take by hand. For external context, professionals expect AI to save up to 12 hours a week by 2029 (Thomson Reuters).

For decades the bottleneck in contracting was reading time. That constraint is gone. When a contract can be reviewed in minutes, the question is no longer whether you can afford to check every agreement, it is why you would not. The delay in getting deals done is not the analysis any more. It is the process and the negotiation around it, and that is the next thing to fix.

Rafie Faruq, Co-Founder and CEO, GenieAI (draft for sign-off)
5 How Britain negotiates

A historical view of negotiation behaviour on the platform.

This chapter draws on GenieAI's clause-level negotiation history, which peaked in 2025 before the feature moved into the newer agent workflow. It is a historical view, not a current-quarter trend.
Most negotiation was light. Of contracts that reached the negotiation stage, the median was settled in a single round and 90% within four. But it was bimodal: 1% saw more than 60 rounds, and the busiest ran to several hundred. UK businesses negotiated harder than the global average: the 90th-percentile UK contract saw five rounds versus four globally, with a longer intense tail (99th percentile 104 versus 63).

Across 37,428 clause-level negotiation decisions recorded by GenieAI, 78% resolved as agreement, roughly five times the rate of outright rejection.

GenieAI UK Contract Review Report 2026.

UK contracts negotiated more intensely than the global average: the 90th-percentile UK contract saw five rounds of clause-level negotiation versus four globally.

UK n = 2,249; global n = 10,682.

6 By sector and size

Which sectors have the vaguest contracts.

Vagueness is not evenly spread. Technology contracts are the most likely to be flagged for vague wording; public-sector contracts the least, reflecting their formalised, template-driven drafting.
Share of issues that are vague wording, by sector
Technology contracts 19% vague, above the 14.6% average; Public Administration lowest at 7.9%.
SectorVague-wording share
Technology19.0%
Energy15.5%
Consultancy14.6%
Legal services13.1%
Finance13.0%
Real estate11.6%
Construction11.4%
Public administration7.9%
Sectors with N >= 50 issues and >= 5 organisations. Baseline 14.6%.
Two cuts we could not produce honestly: risk by named contract type (typed contracts and reviewed contracts are largely different populations), and any breakdown by the contract owner's role (the role field is populated on 0.04% of users). We flag these rather than estimate them.
7 Sources

External data referenced in this report.

Sources

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; all figures are aggregates. The UK cohort is identified by the organisation's registered country. We report only on contracts reviewed on GenieAI and make no claim to represent all UK contracts. Risk and clause categories are derived by keyword classification of AI-generated issue titles and worked-paragraph text and are non-exclusive. AI review time is measured server-side and is analysis time, not end-to-end time to signature; GenieAI does not record signing. Review-time medians are reported for the most recent complete quarter, since usage has shifted between workflows over time.

Sample365,360 contracts from 244,337 organisations (UK cohort: 66,234 contracts from 32,448 organisations). Risk categories from 37,594 issues across 3,692 contracts.
Period2020-06 to 2026-07; headline current-state metrics from the most recent complete quarter.
Data as atJuly 2026
  • No figure is reported for any group below 50 observations (30 for supporting cuts) or fewer than 5 distinct organisations, to prevent any single customer influencing a published number.
  • All figures read from a single database snapshot on 15 July 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). The UK Contract Review Report. https://www.genieai.co/research/uk-contract-review-report-2026

Get the full data and charts

Download the underlying data tables (CSV), an embeddable chart pack, and the PDF edition with an enterprise appendix.

See how your contracts compare

Genie reviews an entire contract in minutes, flags the vague terms that cause disputes, and drafts the fixes. Built for businesses, not law firms.

Try Genie free Book a demo

Used by over 200,000 people, from startups to enterprises.