Tested first.
Written second.
Every guide on Tool Pantry is built around a real test run — real files, measured numbers, described setup. These are the standards we hold ourselves to.
The core rule: no number ships without a test. Every measurement in our guides comes from running the tool on an actual file with a described setup — and since everything runs in your browser, you can reproduce any of them yourself.
How a guide gets made
- Pick the problem from real demand. Guides exist because people actually search for the task — "video too large for Discord", "photo under 50 KB for a form" — not because a keyword looked good.
- Run the test first. Before writing, we run the tool on a concrete test file and record what happens: input size, output size, time taken, and the settings used. The guides you read contain those actual numbers — for example the compression figures on our benchmarks page or the timings on the processing speed page.
- Describe the setup honestly. Test files and environments are stated ("12-second 1080p clip", "Chrome 155 on macOS") so results can be compared and reproduced — and limits are stated too (what the tool can't do, what might fail, what we skipped).
- Verify, then publish. Every tool involved has passed its automated end-to-end tests; every factual claim in the text is checked against the test run. Anything that fails its check gets rewritten or cut.
Independence
- No sponsored placements. Companies cannot pay to appear in a guide, a comparison, or a recommendation. Ad slots on the site are visually separated from content and have no influence on what we write.
- No affiliate-driven advice. We don't earn commissions on recommendations, so the recommendation is just what the test showed.
- Competitors are treated fairly. Our comparison pages (for example vs Smallpdf) include the honest "when you should use them instead" section — if another tool is better for your case, we say so.
About AI assistance
Like most small teams in 2026, we use modern tooling — including AI assistance — for drafting and engineering. The part that matters for trust is the verification gate described above: every number and every factual claim is produced by a real test run and checked before publishing. AI drafts, tests decide.
Corrections and updates
- Report an error: reach us via GitHub. Verified errors are fixed promptly.
- Visible dates: guides carry published and modified dates; benchmark pages are re-run when the tools they describe change meaningfully.
- Platform changes: when a platform changes a limit (like a new attachment cap), affected guides are reviewed and updated — our file size limits reference is the canonical tracker for those.
What we don't do
We don't publish "top 10 tools" listicles ranked by whoever paid, we don't use fake countdown timers or invented urgency, and we don't claim capabilities a tool doesn't have. When a tool has an honest limitation — the video face blur doesn't track moving faces, the HEIC-to-JPEG conversion can grow a file — the guide says so in plain words. That's the standard. Hold us to it.
Can a company pay to be featured or ranked in a guide?
No. Editorial content is independent: recommendations are based on what the tests show, not on commercial arrangements. Advertising slots are visually separated from content, and no advertiser has any input into what the guides say.
How do you verify the numbers published in guides?
The numbers are produced by running the actual tool on an actual test file, and the test setup is described in the guide. Because every tool runs locally in your browser, you can reproduce any published measurement with your own file in about a minute.
Do you use AI to write content?
We use modern tooling — including AI assistance — for drafting and site engineering, the same way most small teams do. What matters for trust is the verification gate: every factual claim and every number in a guide comes from a real test run that is checked before publishing, and anything that fails its check is rewritten or removed.
How do I report an error or request a correction?
Reach us through our GitHub profile (github.com/wu736139669). Verified errors are corrected promptly, and guides carry visible published and modified dates so changes are transparent.
How often are guides updated?
Guides are reviewed when the underlying tools change, when platform limits change (file size caps, format support), or when a correction is reported. Benchmark pages are re-run when the tools they describe get meaningful updates.