PARTLY

As of 13 August 2026, AI can only partly create a searchable knowledge base for your customers.

This still needs a person who signs their name to it.

Can you do it?

15 minutesto a draft.

1 hourto something you’d act on.

Cost, all in£0/month

Skill neededchat-fluent

Who has to check ityou

What the alternative costsA specialist implementation service is an alternative, but no price is provided in the available data.

If this goes wrong: customers rely on an incomplete or outdated answer, contact support with the wrong expectation, and your team has to correct the information and repair the interaction.

What to actually do

  1. Hand it to a person

    The route this page recommends

    A person who owns the outcome does this end to end, worth it when the failure is dear.

  2. Use a tool built for this

    Second choice

  3. Do it yourself

    The distant third

    A chat interface, chat-fluent skill, and roughly 1 hour until you can act on the result.

    How to actually do it

    1. Open a working document and gather your current help articles, product instructions, returns policy, delivery information, price list, contact routes and repeated support answers.
    2. Remove obsolete drafts and label each remaining source with its owner or review date so the model can identify conflicting or potentially outdated content.
    3. Paste the labelled source material into the prompt, replace the business name and describe the customers and products the knowledge base must cover.
    4. Ask the chatbot to produce the structure, articles, FAQs, search terms and unanswered-question list specified in the prompt.
    5. Compare every drafted answer, condition, price, time and contact detail against the current source documents, then correct or remove anything that cannot be supported.
    6. Create a small test set from real customer questions, including misspellings and different ways of asking the same thing, and check that the intended article or answer is returned.
    7. Load the approved content into CustomGPT or another suitable knowledge-base tool, enable citations or source links where available, and test the customer view before publishing it.
    8. Send questions outside the approved content to a named support route, then set a process for updating the source documents and rechecking the published knowledge base.

    Prompt

    Create the content and structure for a searchable customer knowledge base for [BUSINESS NAME]. Use only the source material I paste below. Do not invent products, policies, prices, delivery times, legal terms or contact details. If the sources conflict, mark the conflict and ask me which source is authoritative. If information is missing, write "Not covered in the supplied content" rather than guessing.

Organise the result into:
1. A clear category and article structure.
2. Short help articles with descriptive titles.
3. A customer FAQ with natural search questions and direct answers.
4. Search keywords and alternative phrasings for each article.
5. Links between related articles using the exact article titles.
6. A list of unanswered questions, possible duplicate articles and content that needs a business decision.

Write in plain UK English. Keep each answer focused on one customer problem. Preserve exact figures, conditions and wording where they matter. Separate customer-facing copy from editorial notes. Do not claim that the content is current unless the supplied material says so.

Before finishing, produce a test list of realistic customer searches, the article or answer each should return, and any search for which the knowledge base should refuse to answer or direct the customer to a person.

Source material:
[PASTE YOUR CURRENT HELP ARTICLES, PRODUCT INFORMATION, POLICIES, PRICE LISTS, DELIVERY INFORMATION AND SUPPORT ANSWERS HERE]

    Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.

What it gets wrong

What caps this at PARTLY: context depth, verification cost and stakes of error.

How we scored this

Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.

AxisScore (0–2)
Output2
Inputs1
Verification1
Liability1
Effort delta2
Total7 / 10

The methodology and its thresholds are published in full.

FAQ

Can AI build a knowledge base from my documents?
Yes, it can organise supplied documents into articles, FAQs and searchable answers, and purpose-built tools can make the result available to customers. It cannot fill gaps or resolve conflicting policies without a decision from your business.
Can AI create a customer support chatbot from my website?
Yes, tools such as CustomGPT can create a chatbot from business content and provide citations. You still need to check the source content, test customer questions and define when the chatbot must pass the conversation to a person.
How do I stop an AI knowledge base giving wrong answers?
Give it an approved, current source set and instruct it not to guess when the answer is missing. Test real customer questions against the published result, remove unsupported claims and route uncertain or high-stakes questions to your support team.
Can AI keep my knowledge base up to date?
It can help rewrite and reorganise new material, but it cannot reliably know when your prices, policies or processes have changed. Assign a person to approve updates and retest the customer search after each important change.

Nearby answers

All Help docs & FAQs rulings

Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.

The newsletter

AI news, new answers and product picks, straight to your inbox.