Version 2026-09-21. These statements apply to this website, our sales materials and every report we deliver. The Terms of Service govern if there is any conflict.
Grades and findings
A grade (A to F) or finding is our opinion. It comes from applying our published rubric to the text as it was given to us or as we captured it on the stated date. It is not:
- legal advice;
- a determination by the ASA, CMA, FCA or any other regulator;
- a finding that anyone has broken any law, code or rule.
Our rulesets are built and tested in house. No regulator has approved them.
Third-party brands
We do not name, depict or grade identifiable businesses in our own marketing, examples or public measurements. Examples use fictional brands or anonymised data. Where a customer commissions an audit that analyses a named competitor's public advertising, the report is confidential to that customer, the competitor's trade marks are used only to identify the advertising analysed, and the competitor has not commissioned, endorsed or reviewed it.
AI model names
Where a report names the AI models tested, the names identify the systems tested and are trade marks of their developers. Those developers have not endorsed our reports or services, and we have no relationship with them beyond being an ordinary API customer.
A model's choice in our test is not the developer's opinion.
Agent Preference Audit results
Results describe how the named AI models behaved under our stated, pre-registered method, on the stated dates:
- They are statistical and point-in-time. Model behaviour changes.
- They are not a rating of any product and not a prediction of consumer behaviour.
Where we show what would happen if something changed — for example, the share of decisions a competitor wins if it publishes a machine-readable record — that is a hypothetical scenario built with the offers we tested. It is not an observed market outcome.
If every model on the panel comes from one vendor, the report says so. Results from a single-vendor panel may not generalise to other vendors' models.
Not a financial promotion
Nothing on this site or in our reports invites or induces anyone to enter into a credit agreement or any other regulated financial product. Rates and offers appear only as quoted examples of advertising copy under analysis. They are not recommendations, and we do not rank financial products for consumers.
UIE Ltd is not authorised by the Financial Conduct Authority.
Our own performance figures
Every figure we publish about our own product has a stated basis. Performance on other material may differ. For example:
Our backtest against published ASA rulings was run in house, most recently on 20 September 2026, against 57 published ASA rulings dated January 2024 to September 2026. Of the 36 upheld rulings within our engine's declared scope, it fired the rule matching the ruling's breach in 31 (86.1%). On the 9 in-scope ads the ASA cleared, a rule of the same family still fired on 2 (22.2%), and across all 14 cleared ads at least one blocking finding appeared on 13 (92.9%). So the engine flags more than a regulator would, and a finding is a prompt to check, not a verdict. The sample is small. The backtest is not an endorsement by the ASA.
We keep a dated evidence file for each figure. Where we say we are not aware of a comparable product, that reflects our own market review on the stated date. It is not a guarantee that none exists.
Unmeasured
Where a metric reads UNMEASURED, no measurement exists yet. For example, how often AI agents fetch a Record without being prompted is unmeasured. Please do not read UNMEASURED as zero, or as a forecast.
Right of reply
If you believe our published material identifies your business or says something inaccurate about it, email legal@worldauth.com. We will review it promptly and correct any error.