How we research and evaluate AI tools
We separate documented facts, editorial judgment and hands-on evidence so you can see exactly what supports a recommendation.
Our editorial principles
Independent research
Selection and verdicts are editorial decisions, not sponsored placements.
Transparent pricing status
Unknown or stale prices are masked instead of presented as current facts.
Documented limitations
Weaknesses and evidence gaps appear next to the benefits.
How tools are selected
We cover products that freelancers, consultants, small agencies and service businesses use for recurring work. We prioritize products with clear use cases in writing, productivity, meetings, automation and client operations.
What “tested” means here
Every tool carries an evidence level. Direct hands-on testing requires a documented workflow, result and date. Public free access tested means the free product was used. Official documentation reviewed means vendor sources were checked. Editorial research only means no independent hands-on verification occurred this cycle.
How pricing is handled
A precise price is shown only when the data record contains a source and verification date. Otherwise the interface says “Free plan available” or “Paid plans available” rather than guessing.
How verdicts are decided
We avoid arbitrary scores. A useful verdict names the audience and trade-off: best for solo consultants, strongest for long documents, or easiest free starting point.
Affiliate independence
Affiliate availability never determines inclusion or ranking. The site currently contains no active affiliate links. Any future monetized link will be disclosed on the affected page.
When a source cannot be read
Some vendors block automated access to their own pricing pages. When we cannot reach an official source, we do not fall back to a third-party figure, a cached value or an estimate. The price stays masked and the tool page states that verification failed this cycle. Two tools in the database are currently in that state.
How the data and the writing stay consistent
Every price, free-plan status and evidence level lives in a single structured record per tool. Pages are generated from those records, so a comparison table and a tool page cannot disagree. Our build-time audit additionally rejects the site if editorial copy contradicts the underlying data — for example, if a tool's write-up advertises a free plan the pricing record says does not exist.
What would make us remove a tool
Very little. Removing tools that score badly would quietly turn the database into a recommendation list. A tool leaves only if it is discontinued or if it turns out to be outside the audience this site serves. Poor fit is documented on the page instead, under “skip it if”.
Corrections and updates
Pages carry a review date and evidence status. If you believe something here is wrong, we would rather hear it than not — vendors included. Once the domain and contact mailbox are active, corrections will be handled through the contact page, and material corrections will be noted on the affected page rather than silently edited.