Which tool to choose
For writing and cleaning up the FAQ, any free AI assistant is fine. The difference is made by what you feed it: if you paste your support emails, the AI extracts the recurring questions better than you'd reconstruct them from memory. Claude handles a big block of pasted conversations well; for a few questions any assistant is enough.
If your site runs on a common management system (WordPress, Shopify), there's often already a "FAQ" block or an accordion component to fill in: the AI writes the content, the management system gives it the right form by itself.
How to do it
From browser or app the procedure doesn't change: you're preparing text, not operating on a specific device.
- Fish the questions from real sources: the emails of people who ask before buying, the chats, social messages, and the questions Google shows under "People also ask" when searching for your product.
- Paste everything into the AI and have it extract the recurring questions grouped by theme (shipping, returns, payments, product use).
- Generate the answers with the request below, one at a time or in bulk.
- Check that each answer is true for your business: shipping times, costs, return conditions. Here the AI can't know your numbers, you give them to it.
The operational syntax:
Below I paste real emails and messages from my customers. Extract the questions that come up most often, group them by theme, and for each write me an answer in English. Rules: the first sentence answers directly, then at most two sentences of context. Simple tone, no jargon. Write the questions the way a customer types them, not in bureaucratese. Here are the materials:
[paste emails, chats, messages]
A useful technical note. Well-structured FAQ pages no longer serve only the customer on the site: today they're read by search assistants (the answers generated by Google, by ChatGPT, by Perplexity) that fish precisely the clean question-answer pairs to build their answers. A clear FAQ is the most direct way to get yourself cited when someone asks an AI about you.
A real example
You sell mattresses online. Combing through the emails you find the same question comes up twenty times: "If I don't get on well with it, can I return it after sleeping on it?" It's the doubt that blocks the purchase.
You paste the emails into the AI. It turns it into: "Can I return the mattress after sleeping on it? Yes. You have 100 trial nights: if within that period you don't get on well with it, we collect it for free and refund you. You don't need to repackage it in the original box." Answer in the first word, conditions right after. That single block, placed at the top of the FAQ, dissolves the doubt that was making people close the page.
When it does NOT work (and how to fix it)
If the FAQ seems like filler
Invented questions like "Why choose us?" aren't questions, they're advertising in disguise. The customer skips them and search engines ignore them. Remedy: keep only the questions someone has actually asked you. If no one has asked it, you probably don't need it. Better eight real ones than thirty inflated ones.
If the same FAQ is copied on multiple pages
Putting the same identical block on twenty product pages seems efficient, but duplicate content gets hidden by the engines: you risk all the copies disappearing. Remedy: keep one single general FAQ page, and on the individual product pages put only the two or three questions specific to that product, written on purpose.
If the answers are long and no one reads them
The AI tends to over-explain. A FAQ with ten-line answers gets read neither by customers nor by search assistants. Remedy: impose "first sentence = sharp answer, then at most two sentences of context" and cut the rest. If an answer needs a whole page, it isn't a FAQ, it's a separate article you link to with a real link.
A tip from someone who actually uses it
The best FAQ comes from the objection log, not from imagination. For a month, note every question that comes to you before a sale, even on WhatsApp, even on the phone. Those are the exact sentences that block people, and that's where the FAQ works: not to inform someone already convinced, but to remove the last doubt from someone on the fence. Pass that log to the AI and you'll have a page that sells while you sleep.
Frequently asked questions
How many questions should a good FAQ page have?
Between eight and ten quality question-answer pairs are the sweet spot for most businesses: enough to cover the serious doubts, not so many as to dilute the important answers. If you have many more, split them by thematic page (a shipping FAQ, a product FAQ).
Do I have to add the special code that makes the FAQ appear on Google?
That code is called structured data (a hidden tag that tells the engines "here's a question and here's the answer"). Today it no longer makes a special box appear in the results as it did years ago, but it still helps search assistants read and cite your answers. If your management system adds it by itself with a plugin, keep it; if you have to write it by hand, the AI generates it in a few seconds from your questions.
If I put the answers on the site, will people stop writing to ask?
Partly, and that's the point: every question the FAQ answers is one less email to handle. But the requests don't disappear entirely, and here's the misunderstanding to dispel: a good FAQ doesn't push customers away, it filters the noise and leaves you the time for the real questions, the ones that deserve an answer from you.