Brand & Sentiment
How AI answers frame your brand
Rank, sentiment, position and share of voice for your brand and every competitor you track, broken down by engine and by topic, with the answers behind each score.
Sentiment is scored during every scan, not on request.
Rank in category
#2
Sentiment
Performance matrix · topic × engine
Every
Mention scored for tone
Topic × engine
Performance matrix
Per run
Trend across scans
Any brand
Yours or a competitor's
Your brand in AI answers
Rank, sentiment and position in one place
Brand Insights is the view for the person who owns the story rather than the traffic. It opens on where you rank in your category, how that moved, and how the answers are framing you when they do name you.
- Rank headline with the change since the previous scan
- Sentiment split across every mention, positive, neutral and negative
- Average position: being named fourth is not being named first
Rank in category
#2
Sentiment
Performance matrix · topic × engine
Performance matrix
Sentiment is never uniform
You can be described warmly on one topic and picked apart on another, and it can differ by engine at the same time. The matrix crosses topics against engines so you find the specific cell that is dragging the average down.
- Topic by engine activity, so weak spots are locatable rather than vague
- Metric-switchable trend lines: visibility, sentiment, position or share of voice
- Top rankings by engine and by topic
Benchmarking
Perception only means something in context
Being described as expensive matters only if your competitors are not. Every metric here works for tracked competitors too, on the same axes and from the same runs, so the comparison is genuinely like for like.
- Any tracked brand charted next to yours on the same metric
- Brands are first-class records, so names and colours stay stable over time
- Sentiment compared brand against brand, not against a category average
Region is a real engine parameter, not a filter applied afterwards.
Evidence
Every sentiment score has an answer behind it
Sentiment is scored from the framing in the answers and the citations around them, during stage three of the scan. When a score moves you can open the answers that moved it and read them, rather than trusting a number.
- Scored from the stored answer text, not from a separate model opinion
- Citations captured alongside, so you can see what informed the framing
- Full answer receipts on every data point
best analytics platform for a small ecommerce team
For a small ecommerce team, a few tools stand out depending on what you need most:
Northlane is usually the best starting point.[1] Setup takes minutes and the reporting is built around revenue rather than pageviews.[2]
Everline is a strong alternative if you already run a warehouse,[3] and Portside is worth a look at higher volume.
Cited sources
3Also included
For the team that owns the story
Sentiment tracking that a communications lead can actually take into a meeting.
Sentiment by prompt
Tone at the level of the individual question, so you know which conversation is the problem rather than that there is one.
Mention activity
How often you are being named at all, tracked across runs. Volume and tone are different problems with different fixes.
Which sources shaped it
Cross into the sources module to see the pages cited alongside your mentions. Models repeat what they read.
By market
Sentiment and rank per country, because the framing your buyers get in Berlin is genuinely not the framing they get in Boston.
Movement per run
Rank and sentiment are stamped against each scan, so a shift after a launch or a story is visible on the chart rather than anecdotal.
API access
Pull brand metrics through the project API for a leadership update or a client report, instead of rebuilding them in a slide.
The next question
Once you know the framing, find what caused it
Brand and Sources are two views of the same scan. When the tone shifts, the citations tell you which pages shifted it.
Retrievals
4,234
Read how AI describes you
Run a scan and see your rank, your sentiment split and the answers that produced both.