Ask four AI engines about a company that rebranded: 360 answers, and what they got wrong

The engines found our new site. What they got wrong came from outside it: other companies with our name, and directory profiles that still describe the business we used to be.

We’re Synapse Research Ltd. Until this year our site sold NFT and metaverse services. On 23 September 2026 we relaunched it around answer engine optimization (AEO) and AI engineering. The next day we asked four AI engines 30 questions, three times each. That made 360 answers, and we read every one.

What we asked

The 30 questions come in four groups. Nine discovery and eight comparison questions never name us. They ask who provides a service, or how to choose one. Six brand and seven trust questions do name us, such as “What does synapsereality.io do?” and “Is synapsereality.io a legitimate company?”.

The engines were OpenAI (gpt-5-mini), Gemini (gemini-3.5-flash), Perplexity (sonar) and Claude. We called the first three through their APIs, and they searched the web before every answer. Claude searched in 82 of its 90 and fetched a page without searching in 3 more.

We counted a citation when an answer linked to synapsereality.io or our GitHub, in the text or in its sources.

Asked by name, they found us

The brand and trust questions produced 156 answers. 105 of them cited our site. The 204 answers to questions that didn’t name us cited it once.

So the new site was readable and the engines used it. The trouble was in the other sources they put next to it.

Four other Synapses and a dissolved company

“Who founded Synapse Research and when?” got no citation of our site in 12 answers. Every engine either answered about a different organisation with the same name, or asked which one we meant.

The name is shared. Across all 360 answers, four unrelated organisations called Synapse were each cited 16 to 24 times. One is a biomedical research institute in the Netherlands. Another runs biomedical R&D projects, a third is a global health research hub, and the fourth sells pharma services. The UK company register was cited 16 times, for a dissolved UK company with our name.

“What is Synapse Research Ltd?” cited us in 1 of 3 OpenAI runs. The other two listed several Synapse companies and asked the reader to pick.

Dates got mixed up the same way. An automatic check flagged 8 founding years in the answers. Seven of them belonged to the other companies.

The old business, described as current

41 of the 156 brand and trust answers mentioned NFT, metaverse, blockchain or Web3 work. That is our old site’s positioning. 26 of them called it past, with words like “previously”, “legacy” or “no longer”. About 15 described it as what we do today. Every engine did it at least twice, and Perplexity did it most.

Gemini, “Where is Synapse Research Ltd based?”: “This is a digital marketing, SEO, and blockchain marketing agency.”

One OpenAI answer to “What does synapsereality.io do?” listed our new AEO work next to the old NFT services. It read as if we sold both.

Only one answer in 360 gave a founding year of 2015:

Gemini, “Has Synapse Research published any case studies or client results?”: “A B2B digital marketing, SEO, and growth-hacking agency founded in 2015 that helps startups and enterprises scale.”

We were incorporated in 2018. Claude traced the 2015 to an agency directory profile. 28 brand and trust answers gave 2018, from the company registry, our site or our About page.

The engines also disagreed about headcount, because they quoted LinkedIn, a startup database and an agency directory, and those three don’t match.

The old pages are still in Google’s index. They already redirect to the new ones, so that part fixes itself as Google recrawls. The directory profiles won’t fix themselves.

Who the engines cited when nobody asked for us

In the 204 answers that didn’t name us, we counted how many answers cited each source at least once.

Source type	Answers citing it, of 204
a forum site	25
Google’s developer documentation	20
LinkedIn	15
a freelance marketplace	15
the most-cited agency	15
a blogging platform	13
a second blogging platform	10
a second freelance marketplace	9
an agency directory	9
our own site	1

No agency owned these answers. A forum site and Google’s own docs were each cited more often than any agency.

How much the numbers move

The same day, we asked five of the questions again on the three API engines, three runs each. All 15 question and engine pairs landed within one citation of the first reading. Only one moved: Gemini on a single question, from 0 of 3 to 1 of 3.

So a change of one citation on a single question is noise. A change across a whole group of questions is worth reading.

What it cost

The 270 answers from the three API engines cost $1.39: $0.76 for OpenAI, $0.26 for Gemini and $0.37 for Perplexity. The Claude answers ran on a subscription, with no charge per call. The repeat run added $0.26, so the whole study came to $1.65.

If your company changed what it does

What we took from the 360 answers:

• Fix the profiles the engines quote, before your own pages. For us that’s LinkedIn, the directory listings and the startup database.
• Put the founder, the founding year and one address on your site, in plain text and in your Organization markup. Then an answer about your name has one version to use.
• Check the old URLs redirect in one hop. Then wait for the index to turn over, and measure again.
• Ask the questions that name you, and read the answers. A citation count alone would have scored our old NFT positioning as a success.

Run it on your own company

We published the tool as answer-engine-benchmark, MIT licensed, with the code on GitHub. You fill in your name, domain, category and audience. It asks the same four groups of questions. It keeps every answer with its sources and cost, so each number in a report can be checked.

The full launch-week comparison with our search rankings is in AEO vs SEO: what we measured on our own site.