LinkedIn Markdown compatibility test record, 2026-08-24

This page reproduces the dated record exactly as it was written on 2026-08-24. A record is never edited to match later behavior. When a platform changes, a new dated record is added and this one is kept.

LinkedIn Test Record — 2026-08-24

Verdict

LinkedIn's standard post composer is a plain-text destination with media and structured-post attachments, not a Markdown or rich-text renderer. Raw Markdown markers remain visible. A Markdown link such as [OpenAI](https://openai.com) remained literal and did not create a link preview, while a bare https://openai.com in cleaned plain text generated LinkedIn's normal preview card.

The correct first-version LinkedIn handoff is cleaned plain text with stable paragraph spacing, readable bullets, visible URLs and explicit labels for code or quotes. Long Markdown should be condensed into a focused post or redirected into a separately designed LinkedIn Article workflow; it should not be mechanically truncated at 3,000 characters.

Environment and safety boundary

Authenticated account: existing personal LinkedIn account.
Surface: desktop Create post modal opened from https://www.linkedin.com/feed/.
No post, comment, article, event, poll, document or scheduled item was submitted.
No media or document was uploaded.
All test text was removed and the empty composer was dismissed at the end.
Visible post tools: emoji, media, event, celebration, hiring, poll, document, expert request and scheduling.
No visible heading, bold, italic, quote, code, table, footnote, math or HTML formatting control was found.

Mode-level result

Mode	Result	Measured evidence
Raw Markdown	Fail for formatting; readable as source text	Heading, emphasis, list, link, inline-code and quote markers remained literal
Cleaned plain text	Preferred	Plain heading, paragraphs, `•` bullets, visible URL, `Code:` label and `Quote:` label remained readable
Markdown link	Fail	`[label](URL)` remained literal and did not create a link preview
Bare URL	Pass with destination behavior	The visible URL generated LinkedIn's normal preview card after a short delay
GFM table	Fail	Pipes and separator row remained literal text
Markdown image	Fail	`![alt](URL)` remained literal; no image or preview was created
Standard character limit	Verified	3,000 ASCII characters enabled `Post`; 3,001 disabled it and displayed `-1`
Long URL counting	Verified	A 114-character URL consumed its full raw length rather than X-style 23-character weighting
Emoji/CJK counting	Verified for tested characters	1,500 emoji were accepted and 1,501 rejected as `-2`; 3,000 CJK characters were accepted and 3,001 rejected

Raw Markdown and fragile syntax

The first focused fixture contained a heading, emphasis, two list items, a Markdown link, inline code and a quote. LinkedIn placed the lines into editor paragraphs but did not parse any Markdown semantics.

Construct	Observed behavior	Required output rule
Heading	`# Markdown Test` remained literal	Remove heading markers; preserve a standalone opening line
Bold/italic	`Bold` and `_italic_` remained literal	Strip markers; Unicode mathematical alphabets must remain optional and off by default
Lists	Hyphen markers remained ordinary characters	Normalize to `•`, `–` or numbered lines
Markdown link	Full bracket/parenthesis syntax remained literal; no preview appeared	Emit `label: URL` or a bare URL
Inline code	Backticks remained literal	Remove fences and use a short `Code:` label or indentation
Quote	`>` remained literal	Replace with a `Quote:` label or typographic quotation marks
GFM table	All pipes and separator characters remained visible	Convert compact tables to labeled rows; consider a separately uploaded document/image for complex tables
Markdown image	Full `![alt](URL)` syntax remained visible	Remove syntax and create a media-upload/alt-text checklist

Evidence: raw Markdown composer, table and image syntax.

Paragraphs, links and clean output

The cleaned test used a plain title, blank paragraph separators, • bullets, a bare URL, Code: label and Quote: label. The visual composer preserved the intended spacing and bullet characters.

The bare URL generated a preview card for OpenAI. Preview generation is destination-controlled and may fail for other URLs, so the tool must keep the visible URL even when a preview is expected.

Evidence: clean plain-text composer.

Character counting

LinkedIn Help currently documents a 3,000-character limit for standard posts. The live composer matched that limit and behaved like UTF-16 code-unit counting for the tested input:

Input	UTF-16 length	Code points	Composer state
3,000 ASCII characters	3,000	3,000	Posting enabled
3,001 ASCII characters	3,001	3,001	Posting disabled; `-1` displayed
Prefix + space + 114-character URL	3,000	3,000	Posting enabled
Same URL input + one ASCII character	3,001	3,001	Posting disabled
1,500 `😀` emoji	3,000	1,500	Posting enabled
1,501 `😀` emoji	3,002	1,501	Posting disabled; `-2` displayed
3,000 `中` characters	3,000	3,000	Posting enabled
3,001 `中` characters	3,001	3,001	Posting disabled; `-1` displayed

This should be implemented and regression-tested against live fixtures rather than described generically as “3,000 visible characters.” Complex grapheme clusters and normalization were not separately tested.

Evidence: 3,001-character rejection, https://www.linkedin.com/help/linkedin/answer/a528176.

Native controls and adjacent formats

The current modal exposed media, event, celebration and scheduling controls. Its expanded menu exposed hiring, poll, document and expert-request actions. It exposed no text-formatting toolbar.

The feed also offered a separate Write article destination. LinkedIn Help says standard posts are limited to 3,000 characters and recommends the Publishing Platform for longer content. Article conversion remains outside this test because it is a different editor and output contract.

LinkedIn's current Posts API documentation supports text, images, videos, documents, articles, multi-image posts and polls. Media requires separate asset upload flows. These API facts demonstrate future technical possibilities but do not change the static first-version handoff.

Sources:

Not separately live-tested

The following constructs were not individually pasted during this focused post test: Markdown footnotes, math, Mermaid, Obsidian wiki links/transclusions, raw HTML and heading anchors. The safe initial rule is to resolve or flatten them before LinkedIn output, but their matrix cells should remain U until a focused fixture verifies the exact editor behavior.

Development contract

Rule ID	Trigger	Required transformation	Warning/check
`LI-HEAD-01`	Markdown heading	Remove `#`; keep a standalone opening line	Flag more than one headline-like opening
`LI-EMPH-01`	Bold/italic/strike	Remove Markdown markers	Unicode styling off by default; warn about accessibility/search
`LI-LIST-01`	Ordered/unordered/task list	Normalize to readable numbered or `•` lines	Convert task state to `Done:` / `To do:` when meaningful
`LI-LINK-01`	Markdown link	Emit label plus visible absolute URL	Preserve URL even when a preview card may appear
`LI-CODE-01`	Inline/fenced code	Remove fences; retain short readable code with label/indent	Warn or summarize long code
`LI-TABLE-01`	Small table	Convert rows to compact labeled lines	Show before/after preview
`LI-TABLE-02`	Wide/complex table	Do not flatten silently	Suggest document/image plus accessible text summary
`LI-IMAGE-01`	Markdown image	Remove syntax; extract source, caption and alt text	Produce separate media-upload checklist
`LI-LIMIT-01`	Output near/over limit	Count tested UTF-16 length and display remaining value	Block copy-as-ready above 3,000; offer condensation
`LI-PARA-01`	Paragraphs/hard breaks	Normalize excessive blank lines while preserving readable rhythm	Preview exact plain-text output
`LI-LONG-01`	Source cannot fit a focused post	Condense around one idea	Offer separate article workflow only after article-editor testing

Product implications

The LinkedIn output must use text/plain; raw Markdown and generic rich HTML are the wrong contract.
The first lines deserve special treatment because LinkedIn posts are feed content; the product should preserve a clear opening, not a document-style H1 hierarchy.
The character counter must match observed LinkedIn behavior, including emoji surrogate pairs and raw URL length.
Bare URLs are preferable to Markdown link syntax and may produce preview cards.
Tables, diagrams and Markdown images should become attachment plans with captions and alt text rather than broken inline syntax.
The tool should prepare content for direct paste or later handoff to Buffer/Typefully; scheduling and analytics are not the differentiated job.

Cleanup

All test text was removed. The empty Create post modal was dismissed and the feed returned to its normal Start a post state. No post, draft, article, comment, schedule, poll, document or uploaded media was created.

Screenshots from this test

What this record does not cover

A live result proves the tested surface on the tested date. It does not prove every account tier, editor version, theme, plugin, browser, email client, or API path. The limitations section inside the record above lists what was left untested.