Insights

Customers rate the answers; you fix the content behind the bad ones.

A rating left in the chat window stays with the conversation it came from, so a review session can start at the answers customers marked unhelpful. Read the exchange behind each one, then fix the page that produced it.

Build mine for free All product features

What it does

Built for the questions customers ask before they contact you.

A verdict from the person who asked

Every answer carries a thumbs up / thumbs down control, so the verdict comes from the person who needed the answer, at the moment they read it. Rating happens in the chat window itself, so the verdict comes from the person who actually needed the answer, at the moment they read it — not from someone auditing transcripts a fortnight later. The rating is stored with that exchange in the EU and stays with it, so one left by a midnight visitor in a language nobody on your team reads is still waiting for you on Monday.

Puts the weak answers at the top of the pile

Your dashboard filters conversations by the rating a customer left, including going straight to the exchanges people marked unhelpful. A review session then starts with the answers real customers marked as weak, rather than with a week of chat to wade through first.

Shows the exchange behind the rating

A rating on its own diagnoses nothing, so each one opens into the full conversation: the question in the customer's words, the answer they were given, and the point it stopped being useful. Often the wording is fine and the facts are stale — the assistant was faithfully quoting a page of yours.

Points at a page you can change

A poorly rated answer usually means the source content behind it is thin, out of date or missing. You edit your own page, the sync picks the change up, and the improved version is what the following customer hears. Questions your content could not answer at all surface separately, as knowledge gaps. None of this needs configuring or paying for: ratings collect from the first conversation onwards, on a chat window you can have live in ~10 min, on the Free plan at £0.

Setup

How you switch it on.

Connect your website

Adoomi answers from your own pages, FAQs and approved content, and that is what turns a rating into a job you can actually do. A poor rating traces to a page with your name on it — one you can open and edit this afternoon — rather than to a model whose behaviour you would have to argue with.

Review the answers, and the voice they are in

Nothing reaches a customer until you have approved the wording, so that pass is where you approve what it may say before anyone rates it. Customers rate the whole answer, tone included — and this approval step, not the ratings that follow it, is what keeps the facts right in the first place.

Add one line to your website

One line puts the chat window live, and it carries the rating control from day one: no feedback tool to bolt on afterwards, no survey to write, nothing to configure. Connecting the site, reading the answers and pasting that line is the whole job. The rating control ships with every answer from that moment — nothing extra to add.

Give the review a standing slot

A signal only counts if somebody acts on it, so put a recurring slot in one person's diary: filter to the poorly rated conversations, read the worst of them, fix the page behind each. Nothing chases you to do this and ratings wait in the dashboard for that visit.

Works with

Named integrations — and their limits.

Conversation review

Where a rating is read in context: the same inbox keeps every conversation in full, filters to the badly rated ones and shows you the exchange behind each. Limit: review is read-only — nothing lets you rewrite the answer that was rated badly, or send a correction to the customer who received it.

Knowledge gaps

The sibling signal, covering questions your content could not answer at all rather than answers that were given and disappointed. Limit: neither one writes the fix — a gap tells you a subject is missing, a poor rating tells you an answer landed badly, and the new wording is yours either way.

Knowledge sync

Where the fix actually lands: edit your own page and the sync brings the new version into the assistant's answers, running on a schedule and again whenever a source changes. Limit: nothing improves until your content does and that sync has run, so a rating today is not a better answer today.

Email alerts

Alerts fire on a captured lead, a conversation needing a person and a question your content could not answer — a rating is not one of those events, so a poor rating is something you go and look for in review. Limit: no rating-triggered notifications, so nothing tells you an answer landed badly until someone opens the dashboard.

Worked example

A run of poor ratings on delivery answers, and the page behind them

Your teamWeekly review: filtered straight to the conversations customers rated badly. One subject kept repeating — delivery.

CustomerIf I order today, when would it actually turn up?

AdoomiStandard delivery takes about a week from dispatch, and you'll get an email when your order leaves us. If you need it sooner than that, I can pass your details to the team and someone will confirm what's possible.

Your teamRated unhelpful. Reading it back, the wording was not the problem — the assistant answered exactly what our delivery page says, and that page has not been touched since we changed courier. It was quoting us.

Your teamDelivery page rewritten to match how we actually ship now. The sync picked the change up on its own; nobody retaught the assistant anything.

Your teamThe following week, same question from a different visitor:

CustomerHow quickly do you deliver?

AdoomiMainland orders placed before the daily cut-off go out on a next-day courier, and the times for other destinations are on our delivery page. You'll get a tracking email when it's dispatched, and I can pass your postcode to the team if you'd like a firm date.

Your teamThe following week, no new poor ratings had arrived on delivery questions.

The rating never said what was wrong — it said where to look. Nothing announced it either: the run of poor ratings was found because somebody sat down to filter for them. Reading the conversation is what showed the answer was faithful to a page that had gone stale, and the fix landed on that page rather than on the assistant. Ratings recovering on that subject is how this illustration ends, not an outcome Adoomi promises.

Honest limits

What it doesn't do.

Ratings are opt-in, so read them as a sample

Leaving a rating is voluntary and most customers pass on it, so what you see is the vocal slice of your traffic. A run of poor ratings on one subject is a real signal; a quiet week is not a clean bill of health, and silence from a customer is not the same as a good answer. Read ratings alongside unanswered questions and the conversations themselves.

A review signal, not an analytics suite

There is no deflection rate, no CSAT score, no resolution-rate chart and nothing attributing chats to revenue. What you get is which answers were rated badly and the conversations sitting behind them. If you need a figure for a board pack, this is not where it comes from.

A good rating on a wrong answer is still a wrong answer

Customers rate what felt helpful, not what was true, so a confident wrong answer can collect a good rating. Accuracy comes from the mechanism underneath: the assistant answers only from content you approved, you review what it may say before go-live, and it hands to a person rather than guessing. Ratings are the smoke detector, not the fire door.

The signal says where to look; you decide what to change

Nothing edits a page for you, rewrites an answer or proposes new wording. A poor rating points at a subject; whether the page behind it is wrong, thin, hard to find or perfectly fine is a judgement you make by reading the conversation.

Nothing goes back to the customer who rated

There is no apology email, no follow-up and no notification that the answer has been fixed. The person who rated it badly has gone; the improvement is for the next person who asks the same question.

FAQ

Questions buyers ask about helpfulness signals.

What is a helpfulness signal, exactly?

A helpfulness signal is a rating a customer leaves on an answer in the chat window. The rating stays attached to that conversation, so your dashboard can filter to the badly rated exchanges and you can open any rating to read the question and answer behind it.

What do I actually do with a bad rating?

A poor rating usually points at your own content rather than at the assistant: a page that is thin, out of date or missing. Open the conversation, read what was asked and answered, edit that page, and the next sync puts the corrected version into the assistant's answers.

Will I be told when an answer is rated badly?

Ratings raise no notification. Email alerts fire on a captured lead, a conversation needing a person and a question your content could not answer — never on a rating — so a poor rating is found by filtering for it in review. A standing slot in the week is what makes that reliable.

Do ratings tell me how my assistant is performing overall?

Ratings are opt-in, so they are a sample, never the full picture: plenty of customers never rate anything, and silence from a customer is not the same as a good answer. Adoomi shows no deflection rate, no CSAT score and no resolution charts — the signal points you at conversations to read.

Can a wrong answer still get a good rating?

A wrong answer can certainly be rated well, because customers rate what felt helpful rather than what was true. Accuracy comes from elsewhere: the assistant answers only from content you approved, you review what it may say before go-live, and it hands to a person rather than guessing.

Do ratings cost credits?

Rating an answer costs nothing, and neither does reading the ratings: one credit is one AI answer, so credits are spent when the assistant replies. Plans are Free at £0 a month for 50 credits, Local £8 for 200, Growth £20 for 500 and Scale £75 for 4,000.

Useful next product pages.

Launch

Put this feature behind real customer conversations.

Adoomi starts with your public business knowledge and routes the moments that need a person to your team.

Build mine for free

Healthcare

Channels

Help Center