Jaxon Parrott
BriefWritingFAQMachine RelationsPressAbout
← FAQ

What should a founder do when AI engines disagree about the brand?

When the engines disagree, name what differs: a wrong fact, a different recommendation, a missing answer, or two answers to two questions. Correct the fact on the page you control, and treat a repeated citation pattern as the next source to earn.

Start with the actual disagreement. One answer can name the brand correctly while another recommends a competitor. That is a difference in recommendation. Two answers can state incompatible product facts. That is a conflict of fact. A failed collection is a missing observation. An answer with no attached citation is an answer you can still read for the name and the recommendation. Label what differs, and the action follows the label. Keep the comparison matched: the exact prompt, whether the brand was supplied, the date and window, the engine and product, the collection provider, and any model change. The same words through an API and a consumer app are two products. When the records support a matched comparison, you can call the difference an engine preference. When the records are unmatched, say the difference is still open. For each available answer, record the factual claim, the brand mention, the recommendation, the attached source, and whether the cited page supports the claim. Consulted-source telemetry stays in its own column, separate from displayed citations. Then read the pages: who controls them, what they say, how current the fact is, and whether the page fits this buyer question. If a page you maintain states a wrong specification, correct it and keep the change record. If the source is right and an answer misstates it, keep both as evidence for the engine's feedback route. Leave a correct page correct. If an outside article is wrong, send a sourced correction request. A repeated citation pattern names the pages worth inspecting and the source worth earning next. For different recommendations that are each defensible, ask which buyer decision and which engine product matter to the business, using the customer and sales evidence you have. A single question and a handful of outputs are a sample. Widen it before you call a preference stable. The brief ends as a correction, an investigation, a bounded test, or an explicit open finding. Two releases are two readings. With comparable observations, widen the sample before you change the plan. The public Index is dated aggregate context. AuthorityTech owns the measurement workflow. The kind of disagreement decides the action.

The Index number

This release includes ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews and Perplexity across 36,189 runs with uneven engine coverage by day. That aggregate coverage does not by itself establish stable engine preferences or explain a disagreement about a particular brand.

Published November 5, 2026, updated November 14, 2026. Counts from Machine Relations Index release November 5, 2026, runs collected May 10 to November 5, 2026. Dataset: doi:10.5281/zenodo.23652423.

Definition
Machine Relations
In practice
AuthorityTech on engines citing YouTube and Reddit instead of the brand

Related answers

  • Who founded Bramwell, Inc.?
  • What is AuthorityTech, from the founder who built it?

FAQ

Which engine should I read first?

Read the products your buyers actually use, each result on its own. Weight them with the business evidence you have. A missing observation stays missing, in its own column.

What is the tie-break when two vendor charts disagree?

Compare the retained prompts, answers, windows, engine products, collection routes, and scoring definitions. A matched comparison can be decided. For conflicting facts, verify the primary source. The public Index is the aggregate context beside both charts.

AuthorityTechMachine Relations
XLinkedIn
© 2026 Jaxon Parrott
GlossaryTopicsSkillsArchiveContactPrivacy