Why Your Site Cannot Answer a Question

Somebody lands on your site at nine on a Tuesday evening with one question. Do you work with firms their size. Can you start before March. Does the thing you sell fit alongside the system they already run. They have a specific question, and they have it because they have already done most of the work of deciding.

Your site has forty pages. None of them answers it, because none of them was written for that person on that evening. So they look at the contact form, decide it is not worth the follow up call it will trigger, and close the tab.

You never find out. The visit that mattered leaves no record at all.

The question is always specific

Nobody arrives wanting an overview. Overviews are what you publish when you do not know who is reading. The person who has already shortlisted you wants one fact confirmed, and it is usually one of about fifteen: size, timeline, geography, what it connects to, who actually does the work, what happens when something goes wrong, how long before anything is live, whether you have done this in their sector before.

Those fifteen answers exist inside your business. Anyone on your team could give any of them in a sentence. They are simply not on the page in a form a stranger can retrieve at nine in the evening without asking a human first.

Why the site was built this way

Because a website is a publication, and publications are organised by the publisher. Your navigation reflects how the company sees itself: services, about, case studies, contact. That structure is good for browsing and useless for answering.

Writing a page per question was never realistic either. Fifteen questions across four types of client is sixty near duplicate pages, which search engines treat as thin content and which nobody would maintain past the first quarter. So the sensible decision was always to leave the detail off the site and let the sales conversation carry it.

That worked for as long as the buyer was willing to start the conversation early.

They stopped starting it early

6sense's B2B Buyer Experience Report 2025, published on 12 November 2025, found buyers now reach the point of first contact at roughly 61% of their journey, down from 69% the year before. The same report found that ninety five percent of the time the winning vendor was already on the buyer's shortlist on day one, and that four out of five deals go to the vendor who was the favourite before any contact took place.

Read those two findings together and your site stops being a brochure. It is the room where the shortlist gets made, and it is the one part of the process where nobody from your company is present.

The visit you never see

Analytics will tell you that a session lasted ninety seconds and ended on the services page. It will not tell you the person wanted to know whether you take on work outside Quebec, could not find out, and went to a competitor whose site said so in the second paragraph.

This is the quietest loss in the business. A lost pitch produces a conversation and a reason. A lost tab produces nothing. Most owners have no idea how many of them happen in a month, which is why the site rarely gets the attention the pipeline number would justify.

What AI changed about this

Retrieval. Until recently, answering a specific question out of your own material meant one of two things: a person reading it, or somebody building a decision tree by hand, question by question, and then maintaining it as the business changed. Both cost more than most sites could justify, which is why almost nobody did it and why the ones who tried ended up with a chatbot that could only book a demo.

A model can now read your service pages, your past proposals, your scoping notes and the answers your team has already given a hundred times, and return the one sentence that answers the question actually asked, in the words the visitor used to ask it. Not a search result that hands back a page and leaves the reading to them. An answer.

The second thing that changed is that it can tell you what was asked. Every question a visitor types is a record of demand in their language, not yours. After a month you stop guessing what the site should say next, because you have a list.

What AI will not do for you

It will not invent an answer nobody in your company has ever given. If your team has never actually agreed what your smallest sensible engagement is, no system can tell a visitor, and it should not try. The work of deciding the answers is yours and it usually takes an afternoon.

It should also not pretend to be a person. A visitor who works out halfway through that they have been talking to a machine that implied otherwise does not come back, and that is a worse outcome than the closed tab you started with. Say what it is, and let it hand over the moment the question needs judgement.

It will not replace the meeting either. The point is to arrive at the meeting with the fifteen small questions already settled, so the conversation starts at the part that needs a human.

What to build instead

Write the fifteen answers down first, as sentences, before anything technical happens. Most of the value is in that document and it is worth doing whether or not anything gets built on top of it.

Put them where they can be retrieved rather than browsed, so a visitor asking in their own words gets the sentence rather than a page to read. Keep a log of what gets asked and read it monthly. Give the thing a clear boundary and a clean handover, so a question outside the fifteen reaches a person quickly instead of being answered badly.

Then check whether the site still needs a contact form as its only route in. Most sites are built on the assumption that the visitor will trade their details for an answer. That trade got worse for the visitor every year and they have noticed.

We build these into the sites we develop rather than bolting a chat widget onto the end, because the answers have to come from your material and the handover has to reach a real person in your business. If your site is quiet and you cannot say why, that is usually where to look. See Web Development.

Sources

6sense, B2B Buyer Experience Report 2025, 12 November 2025. 6sense.com