How AI assistants choose between manufacturer, distributor and dealer sources
By Jason MurphyPublished 3 min read
The short answer
How assistants choose between manufacturer, distributor and dealer sources: they have no declared hierarchy, so they weigh agreement, recency and prominence. Each party is authoritative for different facts, and where the weighing disagrees with ownership, the answer comes from the wrong party.
In this article · 6
Three sources, three kinds of authority
A buyer asking about a product is asking three different questions at once, and each is answered best by a different party.
- What is it, exactly? The manufacturer: dimensions, ratings, materials, compliance, the part number and its revision.
- Who carries it, and can I get it? The distributor and the dealer: which branch, which location, what is on hand, what the lead time is.
- Who can sell and install it? The dealer, and the manufacturer's authorization of that dealer.
What an assistant sees
It sees pages and feeds, not roles. A manufacturer spec sheet, a distributor catalog page, a dealer product page and a marketplace listing all say something about the same item. The assistant is not told that the manufacturer is authoritative for the dimension and the dealer for the stock. It infers a ranking from signals available in the text: how many sources agree, how recent each appears, how prominent the domain is, how specific the page is.
Where the ranking breaks
Popularity and ownership diverge in predictable places.
- A distributor's catalog is indexed more widely than the manufacturer's PDF, so a transcription error in the catalog outvotes the engineering source.
- A dealer page quotes a spec from an older revision that has since been superseded. It is more recent than nothing and more prominent than the data sheet.
- A marketplace listing combines the manufacturer's name with another maker's part number.
- Stock and lead time are stated confidently from a feed that last refreshed days ago.
Stating the hierarchy so it can be read
The assistant cannot be asked to learn the hierarchy. The hierarchy has to be published where it can be read: structured data that names the manufacturer as the source of a spec with a revision date, availability fields that carry the time they were observed, and a record that links the dealer's authorization to the manufacturer's own statement of it.
Each party can do its part without waiting for the others. The manufacturer publishes specs software can read with revision and effective dates. The distributor publishes its SKU with a cross-reference to the manufacturer part number and an availability timestamp. The dealer publishes what it carries, where, and the date the page was last confirmed.
Measuring what the assistant currently says
The way to know where the ranking is breaking is to ask the assistants the questions a buyer would ask, against a record of what is true, and read each wrong answer back to its source. Measuring AI answer accuracy describes the method, and tracing a wrong answer to its source describes what to do with each result.
Questions
Does an assistant know the manufacturer is the authority on a specification?
Not unless the sources say so. It infers a ranking from agreement, recency, prominence and specificity, which can favor a widely copied error over the engineering source.
Can a dealer or distributor influence which source is used?
They can make their own facts unambiguous: dated, sourced and cross-referenced to the manufacturer's identifiers. They cannot dictate the ranking.
What should the manufacturer publish?
Specifications with a revision and an effective date, identifiers in standard formats, and each authorized dealer with the lines and services it is authorized for.
Is this about getting recommended?
No. It is about the correctness of the answer. Which supplier an assistant recommends is outside any party's control.
Labs · Research
Research report
Labs · Field notes
- Dealer network data: who decides which fact is right
A dealer network disagrees with itself by design. The fix is deciding, per fact, who wins.
Dealer and channel data3 min read
- Measuring AI answer accuracy against your own records
An answer is accurate or not only relative to a record. Build the record first, then ask.
What AI says and why3 min read
- Trace a wrong AI answer to its source
Every wrong answer came from somewhere. The work is finding the copy that was wrong first.
What AI says and why3 min read
- Part numbers and measurements that disagree across systems
The same item has four identifiers and three kinds of size. Systems agree until a person or a machine has to pick one.
Product data4 min read
- Duplicate business entities across dealer networks
One physical location, three identities. Machines count them as three businesses.
Dealer and channel data3 min read
- Dealer directory data is wrong in more places than one
One dealer is described in six places. No directory is the record, and a correction in one does not travel to the others.
Dealer and channel data4 min read
Where does your own information disagree?
Get Upstream
One email a week: what we asked the AI, what it got wrong, and how we checked.
Our research reaches 17,000+ followers across our channels. Upstream is the weekly version.
Weekly. Unsubscribe anytime. Privacy