Why ChatGPT Names Your Competitors and Not You: The Corroboration Gap

This is the third article in a series on why good Indian business websites go unmentioned: Part 1 covered how a site describes itself, Part 2 whether its products can be read, and this one covers what the rest of the web says about you.

How does an AI assistant decide which companies to recommend?

When a buyer asks for the top companies in a category, the assistant does not start from your website. It starts from the pages that already answer that question: directory lists, association member pages, trade articles, review sites. Your name gets on the shortlist if it is already on those pages, and your website is then read to confirm the details. The more independent sources agree on the same name, location and products, the more willing the assistant is to put your name in front of a buyer.

Your own website is one voice, and it is the voice with the most obvious reason to exaggerate. Every company calls itself a leader. The assistant knows this and discounts it. What it cannot discount is a chamber of commerce member list that includes you, a trade publication that quoted you, a directory entry that matches your address, and a dozen customers who wrote a line about you somewhere public.

This is the mechanism behind a pattern many owners find unfair: a competitor with weaker products and a bigger web presence appears in every answer, while the better company appears in none. The assistant cannot judge quality, so it counts agreement.

Why does being left off an AI shortlist cost more than losing a Google ranking?

A Google results page shows ten links and a buyer can scroll past the first three. An assistant's answer shows three to five names and the buyer contacts one or two of them. There is no page two, and a company that is not on the shortlist was never considered.

For Indian B2B companies the shift is happening inside the purchase department. Buyers have started asking assistants for a first vendor list before issuing an RFQ, and export buyers in Europe and the United States increasingly ask Perplexity for Indian manufacturers of a specific product. A company that is absent from those answers is absent from the RFQ, and no salesperson will ever learn why.

The loss is entirely in new enquiries that never began, which no analytics tool can show.

Gap 1: Why is your own website not enough evidence for an AI assistant?

If the only place on the internet that describes your company is your own website, an assistant has one source and no corroboration. It cannot confirm that you exist as described, and it is unlikely to put its name to a company it cannot confirm. This is a familiar situation for established Indian firms that grew through relationships and never needed a web footprint.

It looks like this. Thirty years in business, a solid customer list, a website built in 2016 and refreshed in 2022, a LinkedIn page with two hundred followers, and no other mention anywhere: no trade press, no association directory, no supplier list on a customer's site, no interview, no case study published by anyone else. To a human in the industry the company is well known. To an assistant it is a single unverified claim.

The irony is that the corroboration often exists offline. The firm is a member of three associations, has won two awards, and supplies companies whose names would impress anyone. None of it is written down anywhere an assistant can read.

Gap 2: Why do inconsistent details across the web make AI doubt you?

When the places that do mention your company disagree with each other, the assistant cannot tell whether they describe one business or several. Two spellings of the name, three phone numbers, an old address on a directory, a different founding year on LinkedIn: each inconsistency lowers the assistant's confidence that these sources are talking about the same company, and the corroboration you do have stops counting.

Part 1 of this series covered this fault on the website itself; off-site it is the same fault with more authors. The gap is very common in India, because most companies have been listed on directories by someone else. A JustDial entry created by a sales agent in 2014. An IndiaMART page set up by a former employee. A Google Business Profile claimed by the previous web agency with the old office address. A trade association list that still uses the pre-incorporation name. Nobody owns the task of keeping these in step, so they drift apart, and every one of them is still live.

The registered name is a particular problem. Government records, GST registration and company filings use the full legal name. The website uses the brand. The directories use whatever the person filling in the form remembered. An assistant that tries to reconcile "Ganguly Polymers Pvt. Ltd.", "Ganguly Polymers" and "GPL Plastics" without help will often give up.

Gap 3: Why do AI assistants ignore customers who never wrote a review?

A satisfied customer who has never written a word about you in public does not exist as far as an assistant is concerned. Reviews, testimonials on third-party sites, ratings on Google, a mention in a customer's own case study: these are the written evidence that other people trust you. Without them, the assistant has only your word.

Indian B2B culture makes this gap worse. Relationships are personal, praise is spoken, and asking a customer to write a review feels like an imposition. Many owners have never asked, and the few reviews that exist were left by employees or friends, and say nothing a buyer would recognise.

The gap is easy to see and slow to close. A competitor with sixty Google reviews and a handful of customer testimonials on a trade site has a body of independent evidence. A better company with none has a claim. The assistant recommends the evidence.

Gap 4: Why do directory and association lists matter to AI assistants?

Industry lists, directory pages, "top suppliers" articles and association member pages are among the sources assistants use most heavily to answer "who are the main companies in this category". Companies appear in those lists mainly because they asked to be included, submitted their details, or paid a modest membership fee. Companies that never asked are absent, and the assistant has nothing to go on.

Owners often dismiss these lists as low-value marketing. For human buyers that was often true. For assistants, a list compiled by a third party that includes your company alongside your competitors is precisely the kind of independent agreement that turns a claim into a fact.

The right lists differ by category. For an engineering exporter the relevant places might be the EEPC member directory, a state industrial association, and two trade publications. For a Kolkata software firm they might be a technology council listing and a couple of review platforms. Knowing which lists an assistant actually reads for your category, and which are noise, is where most of the effort goes. The assistants will tell you, if you ask the right questions and read what they cite, but it takes a morning per category and the answer differs by product line.

How do I check what the rest of the web says about my company?

You can get a rough picture of how the web describes your company with three quick checks. None of them needs a tool or an account.

1. Ask ChatGPT, Perplexity and Gemini "Who are the top [your category] companies in [your city or state]?" Write down the names that appear. Then ask each assistant "What do you know about [your company name]?" and note whether it can describe you at all, and whether it describes you correctly.
2. Search Google for "Your Company Name" -site:yourdomain.com, with the quotation marks. The results are your corroboration. Count them, and check whether the name, address and phone number agree with each other.
3. Search the same way for the two competitors the assistants named. The difference between their results and yours is, roughly, the gap.

That is enough to see whether a gap exists and roughly how large it is. It will not show which of the four gaps is doing the most damage in your category, which off-site sources assistants actually weigh for your kind of business, or which of your existing listings are hurting you because they contradict the others. Mapping that is the part that needs a proper audit.

Why is closing the corroboration gap not an advertising job?

The instinct when a company learns it is invisible is to spend on advertising. Advertising does not close this gap, because assistants do not count advertisements as corroboration. What counts is a body of independent, consistent, written mentions: correct listings in the directories assistants read, membership pages that carry your details, customers who have written something public, trade press that has quoted you, and a website that states its own identity in a form every one of those sources can be matched against.

Most of this is unglamorous work: reconciling names and addresses, claiming and correcting listings, asking customers for a written line, submitting details to the two or three lists that matter and ignoring the twenty that do not. It is slow to build and, once built, hard for a competitor to copy quickly. That is its value.

Questions people ask about AI recommendations and off-site presence

Does the assistant prefer bigger companies?

Not directly. It prefers better-corroborated companies, and larger companies usually have more corroboration. A small firm with consistent listings, real reviews and a couple of trade mentions can appear ahead of a larger firm whose only voice is its own website.

Which directories matter for AI search in India?

It depends on the category. Google Business Profile matters for almost everyone. Beyond that, the useful sources are usually the industry association or export council for your sector, one or two established trade directories, and the review platforms buyers in your category actually use. Most other directories add noise rather than evidence.

My company has a different brand name from its registered name. Is that a problem?

Only if the web does not connect the two. A brand and a legal name can coexist as long as your website, your directory listings and your official profiles state both and state them consistently, so that an assistant can see they describe one company.

How many reviews do I need?

There is no threshold. Assistants weigh consistency and independence more than volume. A dozen real reviews from identifiable customers, spread over a year, are worth more than a hundred left in one week, which Google itself tends to suppress.

Can I fix this without touching the website?

Partly. Correcting listings and gathering reviews happens off-site. But every off-site source is matched against your website, so if the site itself is unclear about who you are, the off-site work has nothing to attach to. The two go together.