A real report: Wellread’s audit of its own site

Last updated 8 October 2026

This is a summary of a real Wellread report: the audit of our own website, trywellread.com, run on 8 October 2026. The numbers are as they came out. We publish it so you can see what an audit says before you run one, and because a company that measures AI visibility should show its own.

The result: AI named Wellread in 0 of 24 answers where buyers asked for options. Our website was 10 days old when this audit ran. This page is our starting line.

What was asked

• 30 questions a buyer of an AI visibility tool might ask, written from our website and checked by us. None contains the name Wellread.
• 12 of them ask for options (“Best software to…”, “How to monitor…”). The other 18 ask how or whether (“Why does…”, “Is it worth…”).
• Each question was asked once to ChatGPT (OpenAI gpt-5.4-mini through its API, with web search on, asked as a buyer in the United States) and once to Gemini (gemini-3.6-flash through its API, with Google Search on).

The numbers

ChatGPT searched the web for all 30 answers. Gemini searched for 20 of 30 and answered the rest from memory. No answer was lost or failed.

One question, followed through

The question. “Best software to track AI search engine recommendations”, asked to Gemini.

The answer began:

Software designed to track how AI engines (like ChatGPT, Perplexity, Gemini, Claude, and Google AI Overviews) mention, cite, and recommend your brand falls into the category of Generative Engine Optimization (GEO) or Answer Engine Optimization (AEO) tracking tools.

How it was counted. The answer recommended Peec AI first and also named Profound, Otterly.ai, LLMrefs, Scrunch AI, Dageno AI. It did not name Wellread. The question asks for options, so this answer is one of the 24, and it counts as “not named”.

What it read. The answer drew on scrunch.com, zealousweb.com, nathanojaokomo.com, upgrowth.in, orchly.ai, hubspot.com. None is our website.

What our own site said. The audit found that our site only partly answers this question. The closest sentence it found was on the home page: “Every figure in the report can be traced to the answers behind it.” That is a claim about our report, not an answer to “which software should I use”.

The fix. Two items in our plan come from answers like this one: write a page that answers the question directly, and get mentioned on the comparison pages the assistants read.

The recheck. The same question, word for word, is asked again in our next audit.

Who AI named instead

The companies named most in the 24 answers where buyers asked for options:

Where the assistants read

The outside sites the answers drew on most. A rival’s own website is left out, since nobody else can get onto it:

• capterra.com, in 5 answers (review site)
• zapier.com, in 5 answers (article)
• beomniscient.com, in 4 answers (article)
• geoptie.com, in 4 answers (article)

What our website doesn’t answer

6 of the 30 questions found no page of ours at all:

• Why is ChatGPT recommending established rivals instead of my startup?
• How do AI search tools decide which B2B software to suggest?
• Why does Perplexity leave my product out of vendor lists?
• How long does it take for AI search engines to index new content?
• Why are buyers using ChatGPT instead of Google for software research?
• Best tools for real time AI search visibility alerts

One of them is the question we exist to answer. When ChatGPT was asked “Why is ChatGPT recommending established rivals instead of my startup?”, it said:

Usually because ChatGPT is optimizing for what it can verify, recognize, and safely recommend rather than for supporting a specific startup.

When asked about us by name

An audit also asks each assistant one question that does name the company: what is it, and what does it offer? Neither described us correctly. ChatGPT described a different company with a similar name, and Gemini mixed that company with ours. Wellread, at trywellread.com, is an AI visibility audit. A new company whose name others share should expect this, and it is why the report shows what each assistant says about you by name.

What we changed after it

1. Dated our pages in the sitemap, so search engines can see when each one changed. From the audit’s plan. Done.
2. Said who we are as structured data, with a link to our real profile elsewhere. From the audit’s plan. Done.
3. Said what Wellread is in plain words on the home page. Not in the plan: reading the answers, we saw the page never said “AI visibility audit” in its visible text. Done.
4. Writing the pages buyers ask about. From the audit’s plan. Started: how to check that AI crawlers can read your site and how to measure whether AI recommends you.
5. Getting mentioned where the assistants read. From the audit’s plan. Not done yet. This is the slow one.

An earlier audit, and what it shows about waiting

We ran the same 30 questions on 6 October 2026. Then, our website served an empty page to any crawler that doesn’t run JavaScript. We fixed that the next day (how). The result on both days was the same: 0 of 24. Making a site readable is necessary, and it does not make an assistant recommend you by itself, least of all within a day.

How to read these numbers

• They describe these 30 questions, asked once each, on one day. Ask again and a few answers will differ.
• They come from the assistants’ APIs with web search on. The ChatGPT and Gemini apps can answer differently.
• Each answer was read, and is kept, up to its first 3,000 characters. 39 of the 60 answers were longer and were cut there.
• An answer that names no company at all is not counted as leaving us out. In this audit there was none on the questions asking for options.
• A later audit that shows a different number shows that the answers changed. It does not prove which of our changes caused it.

How we measure has every definition. To see your own company’s numbers, run a free audit.