A workspace built around one rule

The problem we set out to fix

Making a video means moving between a research tool, a notes app, a document, an image editor and a chat window — and copying things between all of them. By the time a video is published, the research that justified it is in a browser tab nobody will find again, and the three script drafts are called final, final2 and final-real.

PublishBench keeps all of it attached to the video it belongs to. Research becomes a project. The project holds the script, its versions, the transcript, the thumbnail and every piece of metadata generated for it. Nothing is orphaned.

The constraint we took on

The obvious way to build a YouTube research tool is to show impressive numbers: search volume, RPM, projected revenue, a competition rating out of ten. None of those figures are published by YouTube. Every tool showing them is modelling, extrapolating from a different platform, or inventing.

We decided not to. What you get instead is a smaller set of numbers that are actually real, each labelled with where it came from, plus one clearly-marked ranking heuristic whose formula and weights we publish in full. It is less impressive on a screenshot. It is more useful when you are deciding what to make on a Tuesday morning.

What we will not do

• Invent quantitative facts. No search volume, no RPM, no CPM, no revenue projections, no competition score pulled from nowhere.
• Let you pass off someone else’s work as your own. Study any channel you like — but anything written from a source runs a similarity check against it, and the phrase overlap and the longest identical passage are shown to you rather than buried.
• Present a broken integration as working. If a provider is not configured, the feature says so and returns an error. It does not fall back to a plausible-looking fake result.
• Charge you for our failures. Credits are refunded automatically when a generation fails, and cached research is not charged at all. The monthly thumbnail and transcript allowances work the same way: they are counted when a job starts, because two requests at once must not both slip through the check, and a job that ends up failing gives its unit back to the month it was taken from.

How we make money

Subscriptions, and optional credit top-ups. That is the entire model. We do not sell data and we do not run ads. Starter is a permanent account rather than a timed trial: its credits are a one-time grant, and when they are gone everything you have already made stays readable, editable and exportable.

Independence

PublishBench is not affiliated with, endorsed by, or sponsored by YouTube or Google. We use the public YouTube Data API under its terms, and we only access your channel with read-only permission you explicitly grant and can revoke at any time.

How this is run

It is a small operation, which is the honest reason several things on this site are written the way they are: the status page lists limitations by hand rather than through an incident platform. Support and legal enquiries both go through the contact page and reach the same place.