Your likes lie. Your comments don’t.
Likes count how many people saw your post. Comments tell you what they think. We read every single one and give you a report on sentiment, how people see your brand and the conversations going on underneath.
-
Finally! I’ve been waiting for this since September 😍
-
Oh yes, a coffee that tastes like a scented candle. Geniuses.
-
Lovely, but at €6.20 a coffee I won’t be back.
-
Tastes great, but staff haven’t been paid their overtime for three weeks and nobody’s saying a thing.
-
Is the overtime thing true? I’d like to know before I go back.
-
Less sweet this year, much better.
Fictional example. Authors are replaced with pseudonyms before analysis.
We read the comments on your posts on
- X
41% of the comments in the example are about something else.
What the metrics don’t show
You post a product launch. It gets thousands of likes and hundreds of comments. But read them and something else comes up: people using it to complain about poor service, a news story about the company or a decision the post doesn’t even mention.
That parallel conversation is the one that can turn into a crisis, and the one nobody has time to read comment by comment.
-
Likes aren’t opinions. A post with great reach can still have an annoyed audience.
-
Not every comment carries the same weight. One criticism with 800 likes says more than ten compliments nobody replied to.
-
Irony is misleading. “Great, another price rise 👏” is not a positive comment.
-
Hidden storylines keep coming back. The same topic cropping up in dozens of comments is a signal, not noise.
From post to report in four steps
-
1
You tell us which posts to analyse
Your own posts or your competitors’, on Instagram, Facebook, LinkedIn or X. We also work with CSV exports from other tools.
-
2
We collect the post and its comments
Text, image, authors, dates, likes and replies. Before the analysis, authors’ names are replaced with pseudonyms.
-
3
AI reads every comment
One by one, it classifies the sentiment, the emotion and the stance towards your brand, whether the comment is about the post or something else, whether it’s ironic and whether it needs a reply.
-
4
You get the report
With the summary, narratives, risks and recommendations, plus access to the platform to review every comment, filter and export.
We take your brand’s context into account (launches, recent controversies, competitors) to explain where each conversation comes from.
A report you can read in five minutes and check comment by comment
This is what a page of the report looks like. In the margin, what each part is for.
The latte’s return is overshadowed by claims of unpaid staff overtime and anger over prices.
- Net sentiment
- −13
- −31 weighted by likes
- Off-topic
- 41%
- 92% of them negative
- Critical of the brand
- 47%
- 22% favourable
- Need a reply
- 86
- 51 questions
- Negative 44%
- Mixed 9%
- Neutral 16%
- Positive 31%
- Negative 61%
- Mixed 5%
- Neutral 9%
- Positive 25%
- Off-topic
- On topic
Unpaid staff overtime
High risk312 comments use the announcement to claim the company isn’t paying overtime. They account for 42% of the likes and none has had a reply. It coincides with a news story published that same week.
“Staff haven’t been paid their overtime for three weeks and nobody’s saying a thing.”
Hi, thanks for letting us know. We’re looking into September’s overtime payments with our people team, and we’ll message you privately to hear about your situation.
What each part is
- The headlineThe main takeaway in one sentence, for anyone who only has a minute.
- The figuresCalculated from how each comment was classified, not estimated by the AI.
- The real sentimentCounted by comments and weighted by likes, so you can see which opinions resonate most.
- Hidden storylinesConversations that aren’t about the post. Highlighted in yellow so they don’t get lost among the rest.
- Every conclusion, with evidenceWhy it appears, how much weight it carries and the comments it’s based on.
- What to do nextRecommendations and draft replies in your brand’s tone of voice, ready to review.
What every report includes
-
Hidden storylines
Criticism, controversies or complaints the audience brings up even though the post says nothing about them.
-
Executive summary
What the audience thinks, how the post has landed and what’s going on beneath the surface.
-
Stance towards the brand
Kept separate from sentiment: someone can love the ad and still be angry with the company.
-
Emotions and comment types
Joy, anger or mockery; questions, complaints, purchase intent, competitors, spam or insults.
-
Risks and opportunities
Early signs of a crisis with their risk level, and opportunities that emerge from the comments themselves.
-
Unanswered questions
What your audience is asking that nobody has answered: availability, prices, delivery…
-
Coordinated activity
Near-identical comments from different accounts, or profiles that comment over and over again.
-
Everything verifiable
Filter by sentiment, narrative or type, review every comment and export to CSV.
-
Built for sharing
Read-only access for your team, notes on each analysis and a PDF report.
When the comments are about something else
It’s what a quick skim misses. We flag every comment that isn’t about what the post says, group together the ones that talk about the same thing and work out how much weight they carry.
If your audience uses your summer campaign to talk about a shop closure or a labour dispute, you’ll know it in numbers: how many comments, how many likes, whether it’s growing and which comments to reply to first.
Tastes great, but staff haven’t been paid their overtime for three weeks and nobody’s saying a thing.
Is the overtime thing true? I’d like to know before I go back.
Boycott until you pay your people.
- Comments
- 312
- Of the likes
- 42%
- Replied to
- 0
For the people who answer for the brand
-
Brands and marketing leaders
Find out how a campaign really landed, beyond the reach figures.
-
Communications and crisis management
Spot the conversations that could escalate in time, and the arguments behind them.
-
Community managers
Prioritise what to answer and have draft replies ready to go.
-
Strategy and market research
Analyse your competitors’ posts and what their audiences hold against them.
Your data and your audience’s, protected
We analyse public comments, but we handle them with the same care as any personal data.
What the AI model receives
Pseudonym@marta_gpU5@cafe.lover.bcnU12@jordi_rfU13
Pseudonymised authors. Usernames never reach the AI, only a code (U1, U2…).
Servers in the European Union. The platform and reports are hosted in the EU. AI providers only receive the text, with names already removed; if you need it to stay in Europe, we use a provider with EU servers.
You choose the AI provider. We work with Claude (Anthropic), OpenAI, Google Gemini and Mistral, and can use your company’s own AI account if you already have one.
Restricted access. Everyone logs in with their own user account and role. AI provider keys are stored encrypted.
Data processing agreement. We sign the agreement required by the GDPR before we start.
Three ways to work with us
The price depends on the number of posts and how often you need them analysed. Tell us what you need and we’ll send you a proposal.
-
One-off analysis
To understand how a launch, a campaign or a competitor’s post has landed.
- One or more posts, yours or someone else’s
- Full report as a PDF
- Platform access to review every comment
Priced by volume
Ask for a proposal -
Monthly monitoring
To know every month what your audience says about your key posts.
- The posts you choose each month
- Your brand context, set up just once
- Team notes on every analysis
Priced by volume
Ask for a proposal -
Campaign or crisis
For when a conversation heats up and you need to decide what to answer, and to whom.
- Every post analysed for as long as it lasts
- Risks and unanswered questions, prioritised
- Suggested replies in your brand’s tone
Priced by volume
Ask for a proposal
What people usually ask us
Which social networks can you analyse?
Instagram, Facebook, LinkedIn and X. For other platforms, we can work from an export of the comments in CSV or plain text.
Can I analyse my competitors’ posts?
Yes. We analyse public comments on any public post, whether it’s yours or someone else’s.
How many comments can one analysis cover?
Up to 5,000 comments per post. On very viral posts, we prioritise the most-liked comments and those that need a reply, and the statistics are calculated across all of them.
How long does it take?
The analysis itself takes minutes. The overall turnaround depends on how many posts there are and whether we need to set up your brand’s context the first time.
Does it understand other languages and irony?
Yes. It detects the language of each comment and takes irony and sarcasm into account: “great, another price rise” counts as criticism, not praise. The report is delivered in English, Spanish or Catalan, and you can translate it into the other two without running the analysis again.
How reliable is it?
The percentages are calculated from how each comment is classified; the AI doesn’t make them up. Even so, no automatic classification is perfect, which is why every conclusion links to its comments so you can review them one by one.
Can we use our own AI account?
Yes. If your company already works with Anthropic, OpenAI, Google or Mistral, we connect your account and usage is billed directly to it.
How much does it cost?
It depends on the number of posts and how often you need them analysed: a one-off analysis, monthly monitoring, or support during a campaign or a crisis. Tell us what you need and we’ll send you a proposal.
Which post is worrying you?
Tell us which post is worrying you or which campaign you want to understand. We’ll get back to you as soon as we can.