Docs

What PostLens never does

The refusals that define PostLens: no reach forecast, no authorship claim, no rewriting, no LinkedIn automation, one hovered card, saves only on click.

Most LinkedIn tools are defined by what they add. PostLens is defined at least as much by what it refuses to do. These are hard invariants in the code and in the tests — not settings, tiers, or roadmap items. If a page, a post, or a launch comment ever contradicts this list, the list wins.

The list

  • No reach, virality, or impression forecast. The score is an editorial reading of the writing against eight named factors. Nothing in it estimates how many people will see a post, react to it, or how LinkedIn will rank it.
  • No claim about who or what wrote it. The generic/formulaic check is a text-quality cue about templated writing. It is never presented as evidence that a person or a model authored the draft. Not an AI detector.
  • No rewriting. PostLens judges a draft; it does not write, rewrite, humanise, or enhance one. The hints are fixed sentences tied to the weakest factors.
  • No auto-posting, auto-commenting, auto-scrolling, or scraping. Nothing is ever posted, commented, reacted to, followed, or scrolled on your behalf, and no feed or corpus is collected.
  • Reads one hovered card only. The extension and the desktop app read the single post under your cursor, or the composer you are writing in, on a page you opted into. They never enumerate the feed, open “see more”, read messages, or touch cookies.
  • Saves only on an explicit click. Nothing is stored until you press Save. The public check page persists nothing: no draft text, no result, no IP address. The server holds a keyed hash of the draft plus its result (including the audience line) for five minutes so an identical paste is not re-scored; the draft text itself is not kept, and nothing reaches disk or logs.
  • No invented statistics or testimonials. Every number on this site is measured on this build or is the arithmetic of the rubric. There are no user counts, logo walls, quotes, ratings, or timers.
  • Live errors stay errors. If the scoring provider fails, times out, or is rate-limited, you see the error. An offline heuristic is only ever shown under its own demo label, never under the model's name — and the public check page never shows one at all: past the allowance you get the reset time, not a substitute score.

Why refuse

Each refusal removes a way to be wrong about you. A forecast would pretend to know an audience the model never sees. An authorship claim would assert something no text-only check can prove. Automation would put your account on the wrong side of LinkedIn's terms. Storing drafts would turn a writing tool into a liability. Keeping the list short and hard is what lets the score be trusted for the one thing it does.

Questions

Does PostLens tell me how a post will perform?

No. It reads the writing against a rubric and returns a GO / WAIT / IMPROVE verdict with the weakest factors named. It has no engagement model, no audience model, and no impression data. Any reactions in your swipe file are ones you recorded yourself.

Does PostLens detect whether a model wrote my draft?

No. Not an AI detector. The generic/formulaic penalty scores whether the writing reads templated, which a person can produce as easily as a model. PostLens makes no claim about authorship and never labels a draft as written by anyone or anything.

Will it rewrite my draft?

No. It returns a verdict, per-factor levels, and fixed hints for the weakest factors. You do the editing, then check again.

Does it post, comment, or scroll on LinkedIn for me?

No. Nothing is ever posted, commented, reacted to, followed, scrolled, or clicked on your behalf. The opt-in extension reads the one card under your cursor and the composer you type in, and nothing else.

What does the public check page store?

Nothing is written to disk. The draft leaves your browser only to be scored. For five minutes the server holds a keyed hash of the draft plus its result (including the audience line) so an identical paste is not re-scored; the draft text itself is not kept, nothing reaches disk or logs, and no IP is stored. A short-lived in-memory counter, keyed by a one-way hash rather than the address itself, caps how many checks one network address can run per day and is gone on restart.

What happens when the scoring model is down?

You see an error with the provider named. PostLens never substitutes an offline heuristic under the model's name; demo mode exists only where it is labelled demo, and never on the public check page.

Are the numbers on this site real?

Every latency figure is measured on this build and every rubric number is the arithmetic in shared/core.mjs. There are no user counts, testimonials, ratings, or urgency timers, and a build-time copy lint fails the site if that kind of copy is added.

Related: privacy & data, the rubric, and what is Jev?.