Key takeaways
- Brand Knowledge is the Knowledge layer of Brand Intelligence: product specifications, prices, policies and other key facts, each stored with its source and verification status.
- Facts come from documents you upload or sources you connect, and only information a person in charge has reviewed and approved is used.
- Perception and Accuracy check AI answers against this record, and Content Agent writes from it, so every part of flowgen works from the same facts.
- When a needed fact is missing, flowgen asks the person in charge instead of guessing.
The problem: facts live everywhere
Most companies already have the right facts. They are spread across product manuals, spec sheets, price lists, policy documents and press releases, written at different times by different teams. When something changes, one document gets updated and the others do not.
That matters twice in AI search. First, engines repeat whatever version they can read, so an old figure on one of your own pages can end up in an answer. Second, any AI tool that writes marketing content for you has the same choice to make, and without one agreed record it will pick a plausible number rather than the right one.
What Brand Knowledge holds
Brand Knowledge is the third part of Brand Intelligence, after Perception and Accuracy. It manages the key facts about your products and company, each together with where it came from and whether it has been verified.
| Fact type | Example (demo data) | Where AI tends to get it wrong |
|---|---|---|
| Specifications and runtime | LumiClean Pro runs for up to 60 minutes on one charge | An older model's figure repeated for the new one |
| Capacity | Water tank size of a coffee machine | Numbers copied from a reseller listing |
| Price | Current list price | Promotional prices that have ended |
| Warranty | Warranty period by product | Terms from another market or an old policy |
| Certifications and policies | Certifications held, return policy | Claims nobody can trace to a source |
How a fact gets in, and stays right
- Bring in what you already have. Upload product manuals, spec sheets, price lists, policies and press releases, or connect external data sources.
- Review each fact with its source. Collected information is shown next to the document it came from, so the reviewer can check it against the original.
- Approve before use. Only approved information is used in AI marketing. Information you enter or upload goes through the same review and approval.
- Update in place. When something changes, the existing entry is updated rather than duplicated, and the change is tracked, so AI answers that still use the old version can be found.
One record, three users
Brand Knowledge is not a separate archive. The rest of flowgen reads from it.
- Perception analyzes how AI describes your brand and which information it bases that on, so you can see where its picture differs from your own record.
- Accuracy compares specifications, prices, compatibility and policies in AI answers with the verified facts and lists the errors with their sources. It never decides on its own what is true; the facts always come from Brand Knowledge.
- Content Agent writes content from the verified facts and shows the evidence and source behind each sentence, so the person in charge can review and approve it.
An example from the demo workspace
The flowgen site uses LumiHome, a fictional brand, to show the flow. For example, an AI answer states that LumiClean Pro runs for 30 minutes on one charge. Brand Knowledge holds the approved fact: up to 60 minutes. Accuracy flags the claim as wrong and shows the pages the engine relied on, such as a reseller listing that was never updated.
flowgen cannot change an AI answer directly, and it does not claim to. What it can do is show where the wrong figure comes from, help improve the related content, prepare a correction request for external sites, and track whether the answers change afterwards.
Missing facts are asked for, not guessed
Suppose, for example, that a product page draft for LumiBrew Mini needs a warranty period and nothing on record says what it is. Content Agent does not fill in a likely number. Brand Knowledge flags the missing fact and asks the person in charge to confirm it, and the answer goes through review before it appears in any content.
This is slower than guessing for one draft and much faster over a year, because a wrong warranty term published once tends to be repeated by every engine that reads it.
Where to start
- Start with the facts buyers compare: specifications, prices, warranty and the policies that come up in AI answers about your category.
- Upload the most recent official document for each, not every version you have. Older versions are where conflicting figures come from.
- Name a person in charge of approval for each product line, so a missing or changed fact always has someone to ask.
- Then run Accuracy and see which of your facts AI already gets wrong, and send the fixes to Content Agent.