# How Liftli works (machine-readable)

> The six-step loop, in depth. Human-readable version: https://liftli.ai/how-it-works.html. Updated 2026-08-13.

Liftli is a loop, not a text box. It gathers the user's real work, drafts in their extracted voice, waits for a one-tap yes, and learns from what the post actually did.

## 1. Connect once

Liftli lives inside the AI the user already uses — Claude today (a free Claude account is enough), with ChatGPT and Cursor in beta. No dashboard to learn, nothing installed on LinkedIn. Setup is about two minutes: connect Liftli to the client, then pull in a sample of the user's writing so it can extract their voice.

Free tier is a one-time lifetime allowance, no card: 3 posts, 10 comments, 3 idea extractions, and voice extraction.

## 2. It gathers

Most tools start at a blank topic box. Liftli starts at the user's actual week:

- **Voice notes** — a 90-second ramble into Telegram on the drive home is a post seed.
- **Call transcripts** — the sharp thing said in Tuesday's client call, captured instead of lost.
- **GitHub activity** — what was actually shipped.
- **Chats** — the explanation typed to a colleague that deserved a bigger audience.
- **News in the user's lane** — so they can react while a story is still a story.

This is why Liftli drafts do not read like everyone else's: the input is the user's life, not a shared inspiration library.

## 3. It drafts

Every draft is written in a voice extracted from the user's own writing, kept persistently per person, and matched to their strategy — positioning, audience, calendar, what has worked before.

No draft goes straight from idea to screen. Each runs a full **plan → draft → critique → revise** loop first: Liftli plans the angle, writes it, critiques its own work against voice and strategy, and revises. What the user sees is the version that survived that process.

The same engine drafts comments, so presence is participation as well as publishing.

## 4. The user says yes

The pipeline runs on autopilot right up to a one-tap approval gate. For each draft the user does exactly one of three things: approve, tweak, or skip. Nothing publishes without them.

Why the gate is non-negotiable:

- **Feeds punish unsupervised AI.** Detectably-AI content underperforms human writing in most professional niches.
- **The user's name is the asset.** One post they would never have said, published under their name, costs more than a hundred good ones earn.

## 5. It learns

After posting, Liftli logs outcomes, scores engagement, and surfaces warm leads — people whose engagement patterns suggest a conversation is worth having (up to 100 leads on Pro). Then it recalibrates: next week's plan reflects what this week's posts did. Skips and tweaks feed the same memory, so drafts drift toward the user, not away.

That persistent strategy-plus-memory is the difference between a writing tool and a strategist.

## 6. Publishing

Native publishing and scheduling work one way only: through the platforms' official APIs, connected when the user explicitly links an account. Never through browser extensions, scraping, or bots — the category LinkedIn's User Agreement prohibits and has enforced against.

Liftli sits alongside an existing scheduler happily, but does not require one.

## When this loop is not for you

- If you want a hands-off tool that posts while you look away, the approval gate will frustrate you.
- If you do not use an AI assistant day to day, you would be adopting two things at once.
- If you want a standalone scheduling dashboard with queues and team seats, a dedicated execution app is the better buy.

Liftli is for people whose week is already full of good material and whose bottleneck is turning it into consistent, in-their-own-voice presence.