Every line has a source
- Runs in CI in under two minutes README.md
- 4,000 snapshot tests replaced by 300 CHANGELOG.md
For marketers
The team you never got to hire.
You keep the judgment. Propel does the work: research, drafts, pages, decks, and the measurement nobody has time for. It knows the product as well as engineering does, and every line it writes has a source. The words you never want used stay out of every draft.
Specbot in AI answers
Named in about a quarter of answers about spec testing.
ChatGPT31%
Claude22%
Gemini12%
Perplexity27%
Range across 120 answers each · read Monday
Your voice rules
Never says
revolutionary seamless enterprise-ready
Your month, without and with Propel.
You keep the judgment. The research, the drafts and the measuring stop waiting for a free afternoon.
| Job | Without Propel | With Propel |
|---|---|---|
| Knowing the product | You hear what shipped last. | It reads the changelog, so the news reaches you as a draft.How |
| Claims | Legal flags them a day before launch. | Every line traces to a page or file, with the date it was read.How |
| Voice | A style guide few people open. | Words you never use, set once and checked in every draft.How |
| AI search | One screenshot of one answer. | How often 6 AI assistants name you, with a range.How |
| Research | A week of open tabs. | A competitor teardown when you ask for one.How |
| Deliverables | A wall of text to tidy up. | Decks, documents and sheets in your brand, every version kept.How |
Your week, with a team behind it.
You set the angle and make the calls. The drafts, the research and the numbers arrive ready to judge.
- Your part
- 4 moments
- Propel's part
- 5 pieces of work
-
Mon
-
What finished, what waits for you.
-
Set this month's angle
One sentence in the conversation.
-
-
Tue
-
Their pages and changelog, read and compared.
-
Against your own best work.
-
-
Wed
-
Your voice rules already applied.
-
6 assistants, with ranges.
-
-
Thu
-
Every slide sourced.
-
Edit two slides
It keeps every version.
-
-
Fri
-
Every number with its range. A weekly report by email comes later.
-
Example week, times and product. Each Propel step is something it does today.
Your brand, your rules.
What it may say and how far it may go are set by you and enforced by the product, draft by draft.
Limits read from the product, 12 November 2026 Every limit, explainedAlways waits for you
- Hacker News, Reddit and Product Hunt: it prepares, you post
- Anything that spends money
Never
- A claim it cannot trace to a source
- A word on your never-say list
Can earn more freedom
- Articles on your siteup to: It runs on its own
- Pages for AI answersup to: It runs on its own
- Brand postsup to: It runs on its own
- Email to your usersup to: It runs on its own
Each channel starts at "It drafts, you approve" and steps up only on results. One bad claim drops it straight back.
What do marketers ask?
Which AI assistants does Propel measure?
ChatGPT, Claude, Gemini, Perplexity, Grok and Bing, through their official interfaces with web search on. Google's AI Overviews and AI Mode are not measured yet, and Propel says so.
Can I put its numbers in a report?
Yes. Each number comes with a range and the count of answers behind it, so you can show a real change rather than one lucky screenshot.
Will it stay on brand?
It learns your voice from your own site and README. Words you never want used and claims it must never make are set once, and each draft is checked against them.
What can it hand me besides posts?
Strategy documents, launch decks, spreadsheets and landing pages, in your brand, with every version kept.
The same team, from another seat.
- Founders who buildYou built it. Let it sell.
- Studios and agenciesEvery product, finally marketed.
- Sales and growthMore time with people who want to talk.
- CreatorsGet paid to show software you'd use.
The three jobs it does for every seat: Get talked about, Get found, Get customers.