Editorial and Corrections Policy

Last updated: October 27, 2026

ProvisoIQ publishes prompts, playbooks, templates and guidance for people doing legal work with AI. This policy explains how we decide what to publish, how we test it and how we correct it.

Who writes our material

Our material is written by a lawyer with in-house experience and reviewed before publication. We use AI tools when drafting and testing, and a person reviews and edits everything we publish. We write in plain language, task first, and we state limits as clearly as we state benefits.

Neutrality

We do not accept payment to influence which model, tool or vendor we recommend, rank or compare. We describe categories of tools and the trade-offs between them, and our comparisons run the same task on each model with the same instructions. Advertising, where we carry it, is labeled and kept off model comparison pages, and we accept none from AI model providers or legal AI vendors.

How we test Products

Every Product has a test set of invented documents and tasks, including one-sided, incomplete and badly drafted versions. A Product is released only when it passes its test set on every supported model, its limits are stated on its test card, every output tells the user what to check, and nothing in it asks for more information than the task needs.

We retest when a supported model releases a major version, and we record each retest and its result in the public update log at https://provisoiq.com/update-log/. Fixes go to buyers within their update period.

How we time Products

Each test set includes timed runs. A reviewer completes the test tasks with the Product and without it, and the test card records the median difference, the number of runs, the reviewer, the model and the date. We state a time figure only as that measured result, with its basis, and only while the records are on file. Before a Product has timed results, we describe time only by the steps it removes, such as starting from a tested first pass in place of a blank page.

When we quote a published survey, we name its source and present the figure as the survey's finding.

Reader examples

Examples in the mistakes gallery come from readers, rewritten on invented facts, and are credited by the name each contributor chooses or published anonymously. Composite examples we write ourselves are labeled as composites. We do not publish testimonials in invented voices.

Corrections

If you find an error, write to [email protected]. We acknowledge every report, check it, and fix confirmed errors as quickly as we can. Each correction is recorded in the update log with its date and what changed, and a reader who reported it is credited if they wish. Where an error affects a paid Product, buyers who opted in to update notices receive one email describing the fix.

Changes to this policy

We will post any change here with a new "Last updated" date.