Skip to content
ChatLens360
Revenue & customer experience

"I don't know": the chatbot answer that sends buyers to competitors

Some chatbots fail by saying too little. When the answer exists and the bot says "I don't know", buyers go elsewhere. Here is how to fix it.

Illustration of a falling bar chart with coins slipping away for the article ""I don't know": the chatbot answer that sends buyers to competitors", tagged Revenue, Knowledge base, Customer experience.

Most talk about chatbot mistakes focuses on wrong answers. But there is a quieter failure that loses just as many sales: the bot that does not answer at all. When a buyer asks a question your website clearly answers, and the bot replies "I'm sorry, I don't have that information", the buyer does not dig further. They open a competitor's site.

This post explains why chatbots give up on answerable questions, what that costs, and how to find and fix these gaps.

The good "I don't know" and the bad one

To be clear, "I don't know" is sometimes exactly the right answer. A bot that admits uncertainty is far better than one that invents a policy. You want your bot to say it does not know when the information truly is not available.

The problem is the unnecessary "I don't know": the bot refuses or deflects when the answer is right there on your website or in your documentation.

Customer: Does the jacket come in a tall fit? Chatbot: I'm sorry, I don't have information about that. Please contact our support team.

The product page lists tall sizes clearly. The customer, who was ready to buy, now has to send an email and wait. Many will not bother. Another store with a working size guide is one search away.

Why the answer costs you the sale

An unnecessary "I don't know" hurts in several ways:

  • It adds effort at the worst moment. The buyer is close to deciding. Asking them to email support creates delay and friction.
  • It suggests you do not have the thing. Many customers read "I don't know" as "probably not". They assume you do not offer it.
  • It makes the whole site feel unreliable. If the bot does not know your own products, why trust anything else?
  • It pushes work to your team. Every deflection that turns into an email or call is a ticket your bot was supposed to handle.

Illustrative numbers

These figures are made up to show the effect. Use your own.

Say your bot handles 1,500 product questions a month and 6% get an unnecessary "I don't know". That is 90 buyers left without an answer. If one in five would have bought, and your average order is $80, that is 18 lost sales, or $1,440 a month.

Now add the ones who did email support: perhaps 30 extra tickets at 10 minutes each. That is five hours of staff time spent answering questions that were already answered on your website.

Why bots say "I don't know" when they should know

The causes are usually fixable:

  • Content the bot cannot read. Information lives in images, PDFs, tabs, accordions or pages the bot was never given.
  • Different words. The customer says "tall fit", your page says "long length". The bot does not connect the two.
  • Overly strict instructions. A bot told "only answer if you are 100% certain" will refuse a lot of reasonable questions.
  • Scope set too narrow. The bot was set up for support and refuses any sales question as "out of scope".
  • Missing content. Sometimes the information really is not written down anywhere. That is a content gap, not a bot problem.

A checklist for finding knowledge gaps

You can find unnecessary refusals with a simple routine:

  • Pick 20 questions your website clearly answers. Sizes, materials, compatibility, delivery areas, opening hours, plan features.
  • Ask each one in two or three ways, including casual phrasing and synonyms customers actually use.
  • Mark every refusal or deflection and check whether the answer exists on your site.
  • Note where the answer lives. Is it in a PDF, an image, a table, a support article?
  • Check sales questions specifically. Many bots handle "how do I reset my password" well and fail at "which plan is right for me".
  • Review your bot's instructions for rules that make it refuse too easily.

Fixing the gaps

Each type of gap has a different fix:

  1. If the content exists but the bot cannot find it, make sure the page or document is included in the bot's knowledge, and consider adding a clear text version of anything locked in images or downloads.
  2. If customers use different words, add those words to your content. A line like "also known as tall fit" helps people and bots.
  3. If the instructions are too strict, loosen them: answer when the source supports it, and say clearly when it does not.
  4. If the content is missing, write it. Every real customer question without a written answer is a gap on your website too.
  5. Give the bot a useful fallback. When it truly does not know, it should link to the most relevant page or offer a human, not just stop.

There is a fine balance here. Loosening the bot too much leads to made-up answers. See the real cost of a chatbot hallucination for the other side of the problem.

Find the questions your bot should be answering

ChatLens360 generates questions from your own website and documents, so it knows which answers exist. When your chatbot refuses or deflects something your content covers, the report flags it, shows the source page, and suggests what to add or correct in your knowledge base. Run a free chatbot audit to find your gaps.

Put these ideas into practice

Run an automated audit of your own chatbot for free.