Some of the best content ideas from a chatbot arrive as a plain list of things it could not answer. Here are six from an imaginary month at a small kitchen fitting company:
- "do you take away the old units"
- "can we still use the kitchen during the work"
- "how long does it take to fit a kitchen"
- "is there a guarantee"
- "do you do worktops only"
- "can I supply my own appliances"
None of these are exotic. Every kitchen fitter hears them weekly on the phone. Yet the website says nothing about any of them, because websites are written by people who already know the answers. The assistant, which only knows what it has been given, could not answer, said so, and took the visitor's details instead.
That list is worth more than most keyword research. It is a record of what customers wanted to know and could not find, in their own words. The work is deciding what each question deserves.
Group by the decision, not the wording
Raw gap lists are messy. The same underlying question appears as "how long does it take", "how many days will you be here" and "when will it be finished". Treat those as one item.
A useful way to group is by the decision the visitor is trying to make. In the kitchen list above:
- Disruption. How long it takes, whether they can use the kitchen, whether they need to be home.
- Scope. Worktops only, own appliances, removing old units.
- Risk. The guarantee, what happens if something goes wrong.
Grouped like that, six questions become three topics, and each topic suggests a different kind of content.
Page, FAQ entry, policy or nothing
Not every gap deserves a new web page. Some need a sentence. Some need a decision. Some need nothing.
| What the gap looks like | Best response | Kitchen example |
|---|---|---|
| Repeats often, needs more than a paragraph, affects whether people buy | A new page | "What to expect while we fit your kitchen" |
| Repeats, but the answer is one or two sentences | An FAQ entry on an existing page | "Yes, we remove and dispose of old units" |
| You realise you have never actually decided | A policy decision first, then content | Whether customers can supply their own appliances |
| Specific to one service | A paragraph on that service page | Worktop-only jobs, on the worktops page |
| One-off, personal or off-topic | Nothing, or a Q&A pair if it is quick | "Can you recommend a plumber in Leeds" |
| About something you do not offer | A short Q&A pair saying so | "Do you do bathrooms" |
The third row deserves attention. When the assistant cannot answer "can I supply my own appliances", the real reason may be that the answer depends on who took the call. One fitter says yes, another says only if they are a certain brand. Writing a page before settling that just publishes the inconsistency.
When a question earns its own page
A full page is worth writing when three things are true together: the question keeps coming back, the honest answer has several parts, and the answer affects whether someone goes ahead.
"What happens during a kitchen fit" passes all three. People ask it constantly, a good answer covers timings, preparation, mess, noise and access, and uncertainty about disruption is a real reason people delay booking.
"Do you take away old units" fails the second test. It is a one-line answer. Put it in an FAQ and on the main service page.
For more on using chat questions to sharpen the copy you already have, rather than adding pages, see chat questions that improve your website copy.
A worked example of a month's list
Use these as placeholder numbers and swap in your own. Say the unanswered questions report shows 45 entries for the month.
After grouping, you find 14 about disruption and timing, 9 about scope (appliances, worktops, old units), 6 about guarantees, 5 about finance or payment in stages, and 11 that are one-offs, spam or about services you do not offer.
That gives you a clear plan. Disruption at 14 is your new page. Scope at 9 becomes three FAQ entries, one of which needs a policy decision first. Guarantees at 6 get a short section on the main service page, provided the terms are settled. Payment at 5 might be a single FAQ entry, or might wait a month to see if it grows. The 11 one-offs you skim and leave.
Four pieces of work, most of them small, all based on what real visitors asked.
Closing the gap in both places
Publishing a new page fixes the website for people who find it. It does not automatically fix the assistant. After you publish or edit, re-crawl the site so the assistant reads the new content, then ask it the original question in the visitors' own words.
If the answer is still weak, add a Q&A pair using the exact phrasing from the gap list. Visitors often use words your page does not, like "ripping out" instead of "removal".
Next month, check whether those questions have dropped off the list. If they have, the content is working. If a topic keeps appearing after you have covered it, the page may exist but be hard to find, so link to it from the service page too.
What to do with this month's list
Open your unanswered questions report and copy the last month into a document. Group the entries by the decision behind them, then put each group into one row of the table above. Pick the single largest group and write that page or FAQ entry this week.
If a group turns out to be an undecided policy, put it on the agenda for your next team conversation instead. If you need help drafting clear FAQ wording, the free FAQ generator is a reasonable starting point to edit from, and what to do when your chatbot says I don't know covers the in-conversation side of the same gaps. When the gaps come from something you have just launched, updating your assistant for a new service has a routine for that.