
Strategy Done for you by our team. Let's talk.
Every call carries its source, or it is left out.
AI research is known for making things up. Jinn stores each source first, ties every claim to one, checks it, and leaves out what fails.
At a Glance
Every claim carries its receipt.
Strategy grades your brand as it is, writes the documents to fix it, and shows the source behind every claim it makes.
- What you get
42 report types across four depths: a grade first, then the operating documents, the researched landscape, and the senior read over your own finished reports.
- How it’s sourced
Every document reads from a brand record drawn from your own site and catalog, kept signal by signal, including what is missing. A report waits until that record is fresh, within 30 days.
- Why the receipts are real
On the research-backed reports, a claim cannot be stored without a source. The database refuses it, so no step in the code can skip it.
- What it refuses
To guess quietly. A verdict that can’t be read counts as unsupported, and a report below its accuracy bar does not publish.
Every Strategy plan runs one loop: read, log, check, show, then hand over after your yes.
From the first source to the last page

Sources gathered
Cipher files the sources.
Every page the plan reads is filed before a word is written.

The claims ledger
Atlas logs each call with its source.
Each claim goes in with a thread to its source. A claim with no thread is flagged and kept out of the plan.

The sources appendix
Nessie lights the tabs.
The plan ends on a numbered list of its sources, the same list the writer had to use.
The receipts
Sourced, logged, checked, or left out.
Four steps stand between a research-backed document and the reader. Each one is held in the code or the database, not left to a prompt.
- Sources stored
Every document the research reads is saved before a single word of the report is written. The evidence exists first, so a claim can be checked later against the thing it came from.
- Claims logged
The model states its facts as separate claims, each citing a stored source. No citation, no claim: the database will not hold a claim with no source, so the receipts are not a setting anyone can switch off.
- Checked by a second model
A second model reads each claim against the source text and returns a verdict. The report has to clear an accuracy bar of 0.85 supported claims before it publishes. A verdict that errors or can’t be read is scored as unsupported.
- Regulated claims left out
On research-backed reports, a health, legal, financial or age claim that fails its source check is left out. The removal is written to a log the application can only add to.
The other 20 report types are written from your brand record without the research appendix, and each report says which kind it is.

Where each rule is held
| Rule | Held in |
|---|---|
| A research claim cannot be saved with zero sources. The database refuses it. | The database |
| The application can only add to the audit trail. | The application |
The machinery
The receipts are enforced, not promised.
What separates this from a confident chatbot is that the honesty is built in. These are facts about the code and the database, not good intentions.
A claim needs a source to exist
A research claim is stored with the sources it cites, and the database turns away any claim that cites none. The receipt comes before the claim can exist, below the code where it can’t be bypassed.
The checker fails closed
If the second model’s verdict can’t be read, the claim is treated as unsupported and scores nothing. A broken checker lowers the score; it does not wave a claim through.
The audit trail only grows
Retrievals, verdicts and regulated-claim removals go into a log, and the application can only add to it. The record grows; the application never rewrites an entry after the fact.
The appendix is the writer’s only source list
The numbered list a reader sees is the same list the writing model was held to. It cannot cite a number that isn’t on it, so the receipts you check are the receipts the report was built from.
Honest edges
A gap is labeled, not papered over.
Missing and present-but-low are different things, and the reports keep them apart. When the research had no signal, the report says so where the number would have gone. When the signal arrives, the gap fills.
The engine underneath is the same one behind Jinn’s free research tools, not a separate system. Strategy is the packaging: the catalog, the depths and the reader. The sourcing, the checking and the record of where each claim came from belong to the engine, and they behave the same wherever it runs.
Questions
What people ask about the receipts.
Which reports carry receipts?
22 of the 42 report types run the research and checking steps today. Their claims close with a sources and method appendix or carry citations in the body. The rest are written from your brand record without the appendix, and each says which it is.
What happens to health, legal or financial claims?
For brands in regulated categories such as alcohol, cannabis, nicotine, supplements, gambling, lending and crypto, a sensitive claim that fails its source check is removed from the report, and the removal is logged. It is taken out, not reworded into something vaguer.
What if my brand record is out of date?
The report waits. A research-backed report runs only on a record refreshed within 30 days, so a plan is not written from last season’s brand.
Is this a separate engine from your free tools?
No. Strategy is the packaging around the same research engine as Jinn’s free research tools, and the checking works the same in both.
Strategy: For founding brands, done for you by our team. Let's talk.

Strategy: want to see it run on your brand? Let's talk.
- Tell us about your brand.
- A person on our team reads it.
- We read every application and reply within two business days.
Already use Claude or ChatGPT? Connect your AI, free