Editorial policy · Version 2.0 · 2026-09-05
Research methodology and ownership
Who runs this directory
B2Bagents is operated by Gustav. The directory helps business buyers compare AI products by the work they perform, implementation requirements and the evidence available. Editorial contact: hello@b2bagents.org.
How automated research works
Research automation reads public vendor documentation, pricing, trust and legal pages and assembles an attributed brief. Automated checks require a source ledger, dated access records, supported claims and substantive sections before publication. The researcher and original research date appear on each brief. Passing these checks does not establish that a vendor claim is true or that the product has been tested.
What the labels mean
Initial profile means a short market description with limited sourcing. Automatically researched means a synthesis of public sources that passed automated publication checks. Human reviewed is used only when a named person actually reviewed the research. Hands-on tested requires a documented test. These statuses are different, and automated publication never implies human review.
Evidence, uncertainty and corrections
Vendor capability and outcome claims remain attributed to the vendor. Editorial assessments explain fit and the proof a buyer should request; unknown pricing, permissions or implementation details remain unknown. Sources carry their original access dates. Email hello@b2bagents.org with the profile URL, disputed statement and supporting primary source to request a correction. Updates record the changed evidence rather than silently inventing a new verification date.
Ownership and commercial interests
The directory may later earn money from ads, referrals or sponsored placements. Any paid placement or affiliate relationship must be disclosed beside the relevant link or placement. Products owned by the operator must carry an ownership disclosure and meet the same evidence requirements. Payment does not buy a favorable editorial verdict or remove limitations. The directory does not present commercial placement as an independent ranking.
How to use comparisons and scorecards
Comparisons organize public positioning and buyer questions; they are not performance benchmarks. Pilot scorecards are original editorial templates. Fill in targets, sample sizes and stop conditions for your own workflow before testing. A documented buyer-run pilot can establish evidence that public vendor descriptions cannot.