Teach it your business
Your bot isn't answering from a new page
You'll find what is stopping a page being used, and get the bot answering from it.
Before you start
- You can sign in to your Adoomi dashboard
- You have already added the page How to check
- You have one question in mind that only that page answers
About 10 minutes. You’ll do five things:
- ask the bot and note the reply
- find the page in the list
- read its status and badge
- retrain if the page changed
- check how the bot is allowed to answer
Two different things go wrong here, and the reply tells you which one you have. A refusal such as "I couldn't find that in our knowledge base" means the bot has nothing to answer from. A confident answer with the wrong details means it has the page, but an older copy of it. Work out which you have before you change anything.
Start here
Open your live website in a new private window and ask the bot your question.
Write down the reply word for word. It decides which entry below is yours.
In the bot sidebar, under Bot setup, click Train your bot.
The Web pages tab opens first.
Type part of the address into Search pages, then read the row's Status and any badge beside it.
| What you saw | Check this first |
|---|---|
| The page has no row at all | Whether it was ever added, or fell outside the crawl |
| The row reads pending | Nothing yet. Adoomi is still reading the site. |
| The row reads failed | The reason behind · error on that row |
| The row reads ready, but the answer is out of date | Last crawled on that row |
| The row reads ready, and the bot says it cannot find it | How the bot is allowed to answer |
If it didn’t work
The page was never added
- Check
- Type part of the address into Search pages under Website pages. If nothing matches, Adoomi has never read that page.
- Fix
- Next to Mode, choose Single URL, paste the full address of the page, and click Crawl & index. Wait for the row's Status to read ready.
- Resolved when
- The page has a row of its own reading ready, and your question gets a real answer.
The crawl ran, but never reached that page
- Check
- You added the site with Crawl site, other pages came through, and this one did not. A site crawl stops at 100 pages and three links deep, and skips tag, category, search, print, feed and AMP addresses.
- Fix
- Add the page on its own. Next to Mode, choose Single URL, paste the full address, and click Crawl & index. That route fetches exactly the address you paste and skips nothing.
- Resolved when
- The page has a row of its own reading ready, and the bot quotes it.
The row reads failed
- Check
- Hover the · error beside the pill to read the reason. The common one is: Crawl returned no readable content — the site may be blocking automated crawlers. Add content manually or retry.
- Fix
- Click the bin icon on the row, click Confirm delete, then add the address again. If it fails a second time, your site is turning the crawler away. Write the wording into a custom reply instead, so nothing has to be crawled.
- Resolved when
- The row reads ready, or the same answer now comes from your custom reply.
A red or amber badge sits beside the status
- Check
- Adoomi checks every web-page address once a week, on Monday mornings at 06:00 UTC. 404, Redirect, Auth required, Server error, Timeout and Unreachable each mean it could not read the address at that check. Hover the badge for the date.
- Fix
- Open the address yourself in a private window. If the page has moved, add the new address and delete the old row. If it now asks for a login, Adoomi cannot read it, so put the wording in a custom reply.
- Resolved when
- The new row reads ready and the bot answers your question. The old badge clears at the next weekly check.
The page changed after Adoomi read it
- Check
- Read Last crawled on the row. Any time before your edit means the bot is answering from the older copy.
- Fix
- Open the Retrain tab and click Update bot with latest. Wait for the row to go back to ready, then ask again. A re-crawl can reuse a copy of the page taken in the last seven days, so an edit made this week may need a second attempt later.
- Resolved when
- Last crawled shows a time just now, and the answer matches what the page says.
The row is ready, but the bot says it cannot find it
- Check
- In the bot sidebar, under Bot setup, click Bot behaviour, then the Response rules tab. Under Knowledge behaviour, Strict to sources (recommended) means the bot answers only from what you gave it. A question worded very differently from the page falls back instead.
- Fix
- Ask again using words that appear on the page. If that works, the page is fine and the wording was the problem: add the visitor's phrasing as a custom reply so it lands every time. Leave Strict to sources (recommended) on, or the bot starts answering from general knowledge rather than your site.
- Resolved when
- The same question returns the detail from your page instead of the fallback message.
None of these fixed it
- Check
- You have checked the row, its status, its badge, Last crawled and the response rules, and the page still is not used.
- Fix
- Email hello@adoomi.ai with four things: the page address, your bot's name, the question you asked, and what the row's Status reads.
- Resolved when
- A person at Adoomi replies with the cause and what to change.
What this can’t do
- Adoomi cannot read a page behind a login. The crawler carries no password and no cookies, so it sees only what a stranger sees.
- A site crawl stops at 100 pages and three links deep from the address you gave. Pages past those limits are not added — use Single URL for each one.
- A re-crawl can reuse a copy of the page taken in the last seven days, so an edit made this week may not show at once.
- Deleting a row removes those pages from what the bot knows straight away. Adding the address back starts a fresh read.
- Adoomi cannot tell you which pages of your site carry which answer. Compare the reply against the page yourself.