TRINZIK.AI

Content Engines · the social media engine

Social posts that survive a fact-check, because they get one.

The social media engine turns grounded research into ready-to-post copy for X, Facebook, and LinkedIn, written to each platform’s own craft rules and run through a fixed-order guard pipeline that ends in a two-stage fabrication gate. Posts that fail are dropped, not softened, and a person approves every one before it is used.

How is each platform handled?

Each platform has its own craft rules, enforced by the engine.

The rules are a versioned playbook verified against current platform guidance, not folklore. One deliberate rule across all three: posts never carry an external URL, because links crater organic reach. The post earns the attention; your site converts it.

LinkedIn

The hook lives in the first 150 characters, before the fold cuts it off. Bodies run 1,200 to 1,500 characters, built for dwell time: long enough to reward the click, tight enough to hold it.

X

Posts target 100 to 240 weighted characters under the hard 280 cap, measured with the platform's own weighting rules, where a URL costs 23 and some characters count double. A post that runs over is dropped and regenerated, never truncated mid-thought.

Facebook

400 to 700 characters, written neighbor-to-neighbor rather than press-release. Conversational, specific, and shaped for the feed.

Each generation run produces three to five distinct posts per platform, so your team picks from a slate instead of settling for a single take.

How is fabrication prevented?

Five guards, in fixed order. Dropped, never softened.

Every post walks the same line, every time. A post that fails any guard does not get patched up and waved through; it is discarded and the engine tries again.

  1. 01

    Humanizer lint

    A deterministic screen strips the tells of machine writing: the punctuation habits, the filler, the AI vocabulary.

  2. 02

    AI-meta ban

    Nothing that talks about being generated, prompted, or trained survives. The post reads as your brand, not as software.

  3. 03

    Length gate

    Each platform's band is enforced exactly. Out of band means dropped and regenerated, not trimmed into nonsense.

  4. 04

    The numbers gate

    Any figure, written as a digit or spelled out, that does not appear in the source research kills the post outright. This is the hard machine block on invented statistics.

  5. 05

    Second-pass fact-check

    A separate AI pass, hardened against prompt injection, checks every surviving post against the research data itself. Explicitly ungrounded posts are dropped. If the check cannot parse its result, the whole batch fails closed.

If the fact-check itself ever fails to run, the post ships to your queue flagged “not fact-checked”, so a person knows exactly what they are approving.

Who is in control?

Every post is a draft until a person says otherwise.

Generated posts land in a review queue as drafts. Your team edits them in place, approves, uses, or discards each one, and every move is reversible. Approved posts survive regeneration, so a new batch never bulldozes work a person already signed off on. Nothing posts anywhere by itself: the engine writes, your people publish.

A month of social, grounded and approved.

Book a consult and see the engine run on your market: the research it draws from, the slate of posts it produces per platform, and the approval queue your team controls.