What eight AI engines said about Eidomai before we published anything

Eidomai sells one thing: knowing whether AI engines name you, and changing it. So before this site was indexed we ran our own measurement, the same panel, the same eight engines, the same audit every client gets, and here are the numbers. 286 answers across 22 questions. Eidomai named in none of them. Mention rate 0.0 percent, share of voice 0.0 percent. On the two questions about Eidomai by name, 39 factual claims, all 39 wrong.

Zero is the starting line.

We publish it because the claim we make to clients is small and checkable: we shipped this, the measurement moved from that to this. It has to hold for us first.

What the engines said about a company that did not yet exist to them

Two engines said Eidomai does not exist. Grok said no references appear on Nimbletoad's site and that the name may be an unpublished project or a typo. Perplexity could not verify a product called Eidomai by Nimbletoad and offered instead that eidomai is an Ancient Greek verb meaning to appear or to seem, which is where the name came from, so half right.

Two engines invented a product. ChatGPT and Claude both described an AI content and prompt service built on GPT-4 for small businesses, law firms and real estate agents, generating blog posts, emails and proposals. None of that is true. Claude went further and gave the founder a philosophy-of-mind major and a favorite neuroscience book, neither of which are facts about Drew Hendricks. Copilot invented a second company, "Eidomai Solutions", a branding firm that has never existed, and decided Eidomai serves only real estate.

Gemini, Google AI Overviews and AI Mode guessed wrong in the other direction, calling Eidomai an internal tool or a codename rather than the public product it is. One placed Nimbletoad in San Diego; the office is in Carlsbad.

Why this is the useful part

Every one of those claims is the kind a business never sees, because nobody asks an engine about their own company. Every one is the kind a prospective customer repeats. And the pattern is the one we see across the practices we measure: when the sources are thin, an engine fills the gap with the nearest category it knows, or with a confident guess.

The correction plan for each claim is the same plan a client gets. State the facts plainly on pages the engines read, once, with the name attached. Make the facts available as data. Keep every listing consistent with the site. Ask again next month, on every engine, and report what held.

The twenty buyer questions, and who owns them

The panel is not about Eidomai by name. It is what an owner or marketing lead at a medical practice, an advisory firm, a law firm or a winery asks when AI recommends a competitor or gets a fact wrong. Twenty questions, written from the buyer's side. Eighteen were lost to somebody. Two were owned by nobody at all.

The lost ones went to the category's tools (Profound, Peec AI, Otterly.AI, Semrush, Ahrefs) on the questions about tracking and comparison, and to vertical players on the industry questions: DoctorRank for medical practices, Wealthtender, Ambrose Marketing and WealthReach for advisors, NextGen Wine Marketing for wineries, This Is LD for compliance-sensitive industries as a phrase. Each of those is a real competitor and stays on our list.

The two unclaimed questions are the ones we have written first: who fixes wrong information ChatGPT gives about a practice, and why ChatGPT recommends a competitor instead. Nobody in the category answers them directly. They are also the two questions this measurement is about.

What we are shipping, and what we will report

This library is the first batch: one page per question, in the buyer's words, with the facts in prose and as data. The About page, the accuracy page and the pricing page carry the facts the engines got wrong, stated plainly. The listings match the site.

In October the same 22 questions go to the same eight engines. We will publish what moved as trend and share of voice, never as proof of cause, and a change counts only when it holds across enough samples to be more than chance. If a claim flips, it goes in the corrected column with the date. If it holds, it stays open with its plan. Same rules as every client.

The line under every number

API responses approximate but do not exactly match consumer app answers; we report these numbers as trend and share of voice comparisons, never as absolute truth. That applies to our own numbers first.