Paste your post, see an Authenticity Score on a 0 to 100 scale, and read which specific lines in the hook and body carry the AI signal your network will scroll past inside the first two seconds. The score is the headline; the sentence-level colour map is what you actually act on. LinkedIn posts are short, voice-heavy, and the most pattern-sensitive surface on the major platforms, which means small AI tells in the opener matter more than they would on a blog. This is the pre-publish check professionals and creators run before they hit Post: scan the draft, review the hook and body highlights, revise the generic phrases in your own voice, re-scan to verify the score moved above 75, then publish.
This is the routine active LinkedIn posters actually follow before they hit Post on a draft that matters. The score is the entry point. The hook and body highlights are where the work happens. Re-scanning is what closes the loop before the post leaves the composer.
Open the TextSight detector and paste the full draft from the LinkedIn composer. A typical LinkedIn post is 800 to 2,000 characters and the free tier covers 1,500 words per scan, which is enough for the post plus several revisions inside one session. Strip any draft notes or alternate hooks you were going to delete anyway, because the scorer reads the full pasted block and the tells in a deleted alternate hook can still drag the body score down. The scan returns inside a few seconds.
The Authenticity Score runs 0 to 100 where 100 reads fully human to the classifier and 0 reads fully AI. Useful as a summary; not enough on its own. Underneath, the colour map highlights every sentence that tripped a signal. Green sentences passed. Yellow tripped one or two checks. Red tripped three or more. Read the first two lines (the hook, visible in the feed before See more) and the closer together; those carry the comment-rate decision regardless of how long the post runs.
Open your post alongside the highlights and rewrite the flagged sentences before reaching for any tool. The hook carries the most weight because LinkedIn truncates around 1,300 characters and the feed preview shows roughly the first 210 characters on mobile. Replace a templated opener (Three things keep coming up, Hot take but, The reason most teams get this wrong) with a specific moment: a Tuesday meeting, a number from a client deck, a line someone said on a hiring panel. Cut every generic insight, every comma-separated skills triple, and every What this means for you closer. Replace each with a specific observation grounded in the actual story.
Paste the revised post back into the detector and re-scan. Aim for 75 or higher on a feed post, 80 or higher for an established voice your network already pattern-matches against. If a single line still flags red, go back to step 3 for that one line; do not run a Maximum-mode rewrite pass over the whole post because that flattens the specific anchors that make the post work. Then publish through the LinkedIn composer. TextSight does not interact with the LinkedIn algorithm and we make no promises about specific outcomes; we report the score honestly so you can decide if the post is ready.
A number on its own does not tell you whether to publish. These five bands describe what the classifier is seeing, how your network will likely read the post, and what the right next move is at each band.
Strong personal voice. Specific anchors throughout. Rhythm varies across sentences. Reads as if you typed it between meetings rather than asked a model to draft it. Your network will read past the hook, comment threads will warm up inside the first hour, and the feed algorithm will keep showing the post wider through the day. This is the target band for any post where reach matters and for any account where the byline is associated with thought-leadership credibility.
Most readers will not sense AI patterns. The body of the post sounds like you. One or two yellow sentences are usually flagged and they tend to cluster in the opener or the close. If you have time, rewrite the hook to land on a specific moment and replace any What this means for you closer with a single-line statement. Those are the two highest-leverage edits for moving a 75 into the high 80s.
Half the lines sound polished and generic. Most readers will not finish past the hook. Comment rate will stay cold through the first hour and the algorithm will throttle reach accordingly. Edit the red sentences in the hook and body before publishing if the post matters for distribution. Two or three targeted rewrites usually move a 60 into the high 70s without changing the underlying point of the post.
Comments will openly call the post out as ChatGPT output. The account-level signal is worse: subsequent posts inherit some of the audience perception even when they score well, because LinkedIn audiences carry a baseline expectation for the writers they follow. The fix is structural rather than cosmetic. Restructure the post (drop the templated opener, replace generic enthusiasm with specific reasoning, add real anchors) before resubmitting. A single rewrite pass will not move a score in this band into safe territory.
Almost certainly raw or lightly-edited ChatGPT output. The audience response is mute, unfollow, or scroll-past. Account-level distribution drops noticeably for several weeks afterward. Do not publish at this score for any account where personal brand matters. The fix is a complete rewrite from your own thinking about why the post exists, not a quick edit. Use the AI rewriter on the hardest individual sentences after you have rewritten the structure from scratch.
A thought-leadership essay and a one-line hot take are not the same animal. Each format has its own cadence, its own length, and its own AI-tell pattern. Read the score in the context of the format you are publishing rather than chasing one number across the whole feed.
Five hundred to thirteen hundred characters of structured argument with a hook, two or three observations, and a single-line close. The most common AI tell is the Three things keep coming up opener paired with a What this means for you close, sandwiching three bullet-shaped observations in the middle. Score targets sit around 80 because the format gives the classifier enough signal to read confidently. Long essays past the 1,300-char See more line need an extra anchor every 300 characters to keep the back half from drifting back into AI generality.
A specific moment, three to five beats of narrative, and an implicit takeaway. The format scores well when the story is anchored to a real Tuesday, a real meeting, a real client name, or an industry-specific detail when the client is confidential. ChatGPT softens the specifics into generic stand-ins (a founder I spoke with, a team I worked with, many leaders) which is the single biggest AI tell on this format. Replace generic role-nouns with named anchors and the score moves before any other edit.
Under 280 characters, often a single bold claim followed by one sentence of evidence. Short posts score noisily because the classifier has less prose to read, so scan four or five hot takes together as one paste to spot recurring template phrasing across the set. Common AI tells here are the Unpopular opinion or Hot take but opener, and the rhetorical question close. A confident one-line statement reads more human than a question.
Numbered or bullet-led posts that walk through three to seven points. The format flags hard if every bullet uses parallel grammatical structure and one-line length, which is the ChatGPT default for list output. Vary bullet length aggressively (one fragment, one full sentence, one two-line beat), break parallel structure on at least one item, and the format reads human while still scanning cleanly. Heavy bullet density is the second-strongest visual tell on LinkedIn after the templated opener.
Three hundred to seven hundred characters of compressed argument with no bullets, no obvious structure, and a single resonant close. The hardest format for ChatGPT to produce convincingly because the compression demands voice rather than scaffolding. Pre-publish scans on micro-essays often catch the lesson-handoff close (The takeaway is, Here is what I learned) that AI defaults add even when the rest of the post is in-voice. Cut the explicit lesson and trust the reader to extract it.
Free covers several full posts and several revisions per scan window. Active posters running two or more LinkedIn posts a week usually move to Pro. Full details on the pricing page.
Billed $89.88/year — Save $30
Billed $179.88/year — Save $60
Billed $359.88/year — Save $120
Yearly billing saves 25%. View full pricing →
LinkedIn has not announced a server-side AI classifier on organic posts. The penalty is indirect and arguably stronger for that reason. Here is what the algorithm actually reads and how it translates into reach loss on AI-flavoured posts.
LinkedIn's feed ranking weights how long readers spend on a post as a primary quality signal. AI-flavoured posts collect impressions but lose readers inside the first paragraph because the templated opener tells the audience nothing fresh is coming. Average dwell-time on a typical post is 7 to 12 seconds; AI-flavoured posts drop to 2 to 4. The algorithm reads the gap and stops showing the post wider, even to existing followers.
The first 60 minutes is the window the algorithm uses to decide whether to expand distribution beyond your immediate network. AI-flavoured posts collect roughly half to a third of the comment rate of the same author's voice-driven posts in that window. Once the first hour passes with low comments the post is effectively done; the algorithm rarely re-opens distribution later in the day. The damage is reputational and algorithmic at the same time.
Saves are the slow-burn signal that compounds across weeks. AI-flavoured posts collect very few saves because there is nothing the reader plans to return to. Save rate is also one of the inputs the algorithm uses to predict whether to show your future posts to the same audience. One AI-flavoured post is a slow drag on the next ten because the audience-quality signal for your byline edges down by a small amount each time.
Browser extensions that score LinkedIn post copy for AI-likelihood appeared in 2025, mostly aimed at recruiters evaluating candidate thought-leadership. Adoption is niche but growing. A high AI-score across the last ten posts weakens the credibility signal LinkedIn content is supposed to provide during a hiring evaluation. The score never reaches the candidate, which is the part that makes pre-publish scanning load-bearing rather than optional.
Five tells cover roughly 80 percent of the AI-flavour on posts that readers visibly scroll past. The fix in every case is replacing a templated pattern with one specific anchor. The opener is weighted heaviest because LinkedIn shows it before See more.
Three things keep coming up in my conversations this week, Hot take but, The reason most teams get this wrong is, Recently I had the privilege of, I have been thinking a lot about. ChatGPT cycles through roughly five LinkedIn opener templates and the audience has learned all of them. The opener decides whether readers stay past line two, which is the line where engagement is won or lost. Replace the templated framing with a specific moment: a meeting last Tuesday, a number from a client deck, a line someone said on a hiring panel.
ChatGPT bullet output defaults to three or four items, each a single line, each sharing the same grammatical structure (verb-object, noun-phrase, gerund-clause). Real LinkedIn writers use very few bullets in a 500-word post and when they do use them the lengths vary aggressively. Collapse bullets back to prose where the rhythm allows, keep at most one short list per post, and break parallel structure on at least one item.
A client of mine, a founder I spoke with, a team I worked with, many leaders. ChatGPT prefers generic role-nouns to specific names. Real LinkedIn posts that earn comments name the company, the dollar amount, the date, the exact title. Specificity is what makes a post feel like reporting from inside the market rather than advice from above it. Add at least one verifiable anchor per 200 characters and the comment rate moves measurably.
What this means for you, The takeaway, Here is what I learned, Here is what to do about it. ChatGPT signs off most LinkedIn posts with an explicit lesson handoff. Human writers usually let the lesson sit implicit and trust the reader to extract it. Delete the explicit lesson and close on a concrete detail from the story or a single-line statement that lands like a punchline. A confident close outperforms a hand-holding close on every measurable engagement axis.
In today's fast-paced world, Communication is key, Authenticity matters, Build genuine connections. Generic insight phrasing reads as filler because it carries no information the reader could not have predicted before reading the post. Strip every generic insight, replace each with a specific observation, and the post stops sounding like advice the audience has already read ten times this week.
An honest pre-publish check is closer to a careful proofread than to anything else. We want to be explicit about which side of the LinkedIn-credibility line this scorer sits on so you can decide whether it fits your situation.
LinkedIn posts you wrote yourself, including posts where you used ChatGPT for outline or polish on a draft you wrote. The thinking is yours, the moments are real, the anchors are real. The scorer catches sentences where assistant register leaked into the prose so the published post reads in your own voice. We score honestly so you can decide what the post needs before it goes live in front of your network.
It is not a tool for fabricating expertise or pretending you wrote something you did not. The AI rewriter cannot put authentic insight into a draft whose underlying claims are borrowed wholesale from a template. If your post sounds AI because the perspective itself is generic, the scorer will tell you that and no rewrite pass will magically fix it. The most useful thing TextSight can do for that case is point you back to writing about a moment you actually lived through.
The output of a good revision pass should pass a simple test: if a colleague who follows your feed asked you in person to talk for two minutes about the specific anchor or moment in your post, you should be able to do it confidently. If you cannot, the revision added voice but not substance, and the post will fail the eye-test for the audience members who matter most. The score is a draft check, not a substance check.
Fix posts that scored Mixed or worse with Light, Balanced, and Maximum modes for the feed.
Open the guide →Headline and About section, calibrated for the static part of the profile rather than the feed.
Open the guide →The sister scorer for LinkedIn newsletters and blog articles, calibrated for long-form prose.
Open the scorer →How the 0 to 100 score is computed and what threshold to aim for across formats and surfaces.
Read the guide →Authenticity Score, sentence-level hook and body highlights, 1,300-char feed calibration. Free for several posts per scan window, no signup for first scan.