Four Faults That Make a Well-Designed Website Invisible to ChatGPT

This is the first of three articles on why good Indian business websites go unmentioned; this one covers the faults on the site itself, and how to check for them in about ten minutes.

Why does a good-looking website still get ignored by AI assistants?

Because an assistant reads the page as text and checks four things a design review never looks at: who owns the site, whether it uses one name, whether its headings ask anything, and whether its pages answer anything. A visitor sees layout, photographs and brand. An assistant reads the page as text, tries to work out who is speaking and what they are claiming, and then decides whether that claim is safe to repeat to a buyer. A site can score full marks with the visitor and fail every one of the assistant's tests.

This is why so many owners are caught out. Their designer did a fine job. Their SEO agency got them onto page one for a few terms. Traffic still arrives from Google. Nothing in their reports says anything is wrong, because none of those reports measure what an assistant does with the site.

I learnt this on my own website. During an audit of our own properties in August 2026, I found that the main site named one company as the owner of the business while the blog on the same domain named another. It had been live that way for years. No visitor ever noticed, and no client ever mentioned it. Any assistant that read both pages had two owners to choose from and no reason to trust either.

What happens when ChatGPT recommends a competitor instead of you?

The cost is an enquiry that never reaches you and never shows up in any report. When a purchase manager asks an assistant for a supplier of a specific product in a specific city, the assistant produces a shortlist of three to five names. The buyer contacts one or two of them. If your company was not on that list, you were not rejected. You were never compared.

For an Indian B2B company this matters more every quarter. Purchase teams have started opening ChatGPT or Gemini before a directory, and export buyers increasingly ask Perplexity for a first list of Indian suppliers. Google's own results in India now often open with an AI Overview, and AI Mode has been available here since 2025. Both are built from the same web pages the assistants read.

The frustrating part is that your Google traffic can hold steady while this happens. You keep receiving the enquiries Google sends, so nothing looks wrong. The enquiries that started in an assistant simply go to the company the assistant could understand.

Fault 1: Why does an AI assistant need one sentence that says who you are?

Very few business websites contain a single sentence that states who the company is, where it is based and what it does. The facts exist, but they are scattered: the registered name sits in the footer, the location on the contact page, the product range in a PDF brochure, and the founding year in the About page's third paragraph. A human assembles the picture without noticing. An assistant can assemble it too, but it is far more likely to repeat a sentence that already exists than to compose one and attach your name to it.

The test is simple. Open your home page and look for one sentence that a stranger could read aloud to describe your company completely. On most sites that sentence does not exist. The home page opens with a slogan ("Engineering Excellence Since 1987"), the About page opens with a philosophy ("We believe in building relationships"), and nowhere does anyone say "We manufacture industrial fasteners in Howrah and supply automotive and railway customers across India."

The machine-readable label behind the page, which developers call structured data, makes this worse. The label is supposed to state the organisation's name, address, founding date and official profiles in a form machines can read without guessing. On many Indian business sites it is missing, or it was generated by a plugin that filled in "My WordPress Site" as the organisation name and nobody looked again.

Fault 2: Why do two names for one business confuse AI assistants?

When a website uses two or three different names for the same business, an assistant may treat them as two or three businesses, credit your products to the wrong one, or leave you out because it cannot say which is which. It is one of the first things I look for in an audit, and it is almost never deliberate.

It creeps in through ordinary business life. The company is registered as "Sharma Engineering Works Pvt. Ltd." The brand on the product is "SEW". The blog was set up by an agency that credited "Sharma Group". The Google Business Profile says "Sharma Engineering". The LinkedIn page says "SEW India". Every one of those is correct to a human. To an assistant they are five candidates for the same claim, and it has no proof that they are one company.

The version I found on my own site, a main site and a blog naming different owning companies, is the same fault at a deeper level. The page that should have settled the question of ownership instead contradicted the home page. Any assistant that read both had two owners to choose from.

Fault 3: Why do headings like "Our Solutions" keep you out of AI answers?

Assistants look for passages that answer the buyer's question, and a heading that poses the question with the answer beneath it is the easiest passage to find and to quote. A heading like "Our Solutions", "Why Choose Us" or "Product Range" poses nothing. The buyer asked "Which company supplies food-grade conveyor belts in eastern India?" and nothing on your page comes close, so your page is rarely the one quoted.

This has nothing to do with stuffing keywords into headings. It is about writing headings the way a buyer talks. "Which industries do you supply?" is a heading a buyer might type. "Industries Served" is a heading a designer might choose. The first invites an answer beneath it. The second invites a list of logos.

Look at the headings on your most important service or product page. Count how many of them are questions a buyer would ask aloud. On most sites the answer is zero, and the entire page is a set of labels with paragraphs beneath them.

Fault 4: Why do AI assistants skip pages that never state a fact?

An assistant lifts answers; it does not summarise brochures. If a page carries three paragraphs about the importance of quality in transformer manufacturing but never states what you make, in what capacity range, for which customers, at what lead time, there is no sentence for the assistant to quote with your name attached. It moves on to a competitor whose page said "We manufacture distribution transformers from 25 kVA to 2500 kVA with a standard lead time of four weeks."

The fix in principle is to put the answer in the first sentence beneath each heading and the explanation after it. Most pages do the reverse: context first, answer buried, and often no answer at all because the writer assumed the reader would phone.

Indian B2B sites suffer from this more than most because the sales culture expects a call. The site is written as an invitation to talk rather than as a source of facts. That worked when the visitor was a person. It fails when the visitor is an assistant that will not phone and will not wait.

How can you check your own website for these four faults in ten minutes?

You can run a rough version of this check yourself without any tools. Ask ChatGPT, Perplexity and Gemini the same three questions and compare what comes back with what your website actually says.

1. Ask each assistant "Who owns [yourdomain.com] and what does the company do?" Use the version that searches the web; if it answers without searching, tell it to. If the answer names a different company, a partner, a distributor, or says it cannot tell, you have Fault 1 or Fault 2 or both.
2. Ask each assistant "Which companies supply [your main product] in [your city]?" If competitors appear and you do not, note who appears instead. Their pages are the ones the assistant could read, or the ones the rest of the web vouches for. Part 3 of this series covers the second reason.
3. Open your main product page and read only the headings. If none of them is a question a buyer would ask, you have Fault 3. Now read the first sentence under each heading. If it does not answer the heading, you have Fault 4.

This tells you whether you have a problem. It does not tell you how bad each fault is, which of the four is doing the most damage, or in what order to repair them. That ordering is different for every site, and getting it wrong means fixing the wrong thing first and waiting a quarter to find out. It is the reason the audit exists.

Why does fixing this need more than a content rewrite?

Two of the four faults live in the words on the page. The other two live partly in the words and partly in what the page tells machines about itself, and in the profiles you hold elsewhere. Rewriting the headings and answers fixes the first pair. Naming the business once and correctly, and making every page and every off-site profile agree, needs changes to the template, the labels behind the page and the profiles you hold elsewhere. A copywriter cannot do it alone and a developer will not know it needs doing.

That is why we treat this as one piece of work rather than a content job. The audit establishes which faults are present and how they interact. The repair then goes through the site in one pass: identity, names, headings, answers, in the right order for that site.

Questions people ask about AI assistants and website visibility

Does ranking well on Google mean ChatGPT will mention my business?

No. Google rankings and AI mentions overlap but depend on different things. A site can rank on page one for its main terms and still be absent from every assistant's answer, because an assistant has to be able to read, understand and quote the page, and a ranking does not guarantee any of that.

Which faults are the most common on Indian business websites?

Multiple names for one business and headings that do not ask questions are the two that turn up most often. Both come from ordinary practice: brand names differ from registered names, and design templates use label-style headings by default.

Can my existing web team fix these faults?

Yes, given a clear diagnosis. The audit report lists every fault in priority order with the fix stated plainly enough for any competent developer or copywriter. It is written so that your own team can implement it, or we can.

How long does it take for AI assistants to notice the changes?

Assistants that search the live web, such as Perplexity and ChatGPT search, pick up changes within days to a few weeks. Answers that rely on older training data change more slowly. Plan for a quarter before judging the result, and re-run the same three questions every month to track it.

Is this the same as SEO?

It builds on SEO but is not the same. Good SEO gets a page found and ranked. Answer engine optimisation gets the page understood and quoted. A site needs both, and many Indian SMEs have invested in the first without touching the second.