Back to Insight
Product news
Sentiment analysis: how is AI search talking about your brand?
Andy Francos, Co-founder & CPO
SUMMARY
Sentiment Analysis Plus is now live for every Obsero customer. See how AI search talks about your brand, which platforms and publishers are shaping it, and the attributes driving the score, shown here against Pizza Hut.
Being mentioned in AI search matters to your business, but how you're mentioned matters arguably even more.
A 50% share of voice in your category sounds impressive. If the answers behind that number are neutral or negative about your products and services, though, the first things I'd want to know are which LLM is driving it and which sources are feeding it.
That's why today we're launching Sentiment Analysis Plus in Obsero. It's available to every customer, and it lets you assess the sentiment of every answer in the platform. To show you how it works, we've run it against Pizza Hut over the last 30 days.
Filter to the view you need
When you log in you can filter by topic, by LLM, by prompt type (branded or category) and by market. Date ranges cover the most recent pull, last week, last month or a custom range of your choosing.
The sentiment barometer
The first thing you'll see is the sentiment barometer. It breaks every answer that mentions your brand into segments, scores each one as positive, neutral or negative, and rolls them up into a single score out of 100 that's tracked over time. Next to the score you get the split of positive, neutral and negative.
For Pizza Hut over the last 30 days that score is 70, built from 24,684 scored segments: 11,908 positive, 10,917 neutral and 1,859 negative. The summary underneath picks out the extremes as well. "Best pizza delivery for a kids' birthday party?" scores 88, while "Recommend a good independent pizza place near me instead of a chain" sits at 27, which is hardly a surprise given what that prompt is asking for.
Worth flagging that we only assess sentiment where the brand is actually mentioned in the answer, so answers that leave you out aren't scored.
Sentiment by platform
The platform breakdown pinpoints exactly which LLM is coming back positive, neutral or negative, which tells you where to focus your effort.
ChatGPT and Gemini lead for Pizza Hut on 72, with Google AI Overviews and Copilot on 71 and Perplexity on 68. Perplexity is the biggest climber, up six points on the previous 30 days, while Google AI Overviews has dropped five.
Google AI Mode is the outlier. It scores 64, down 11 points on the previous 30 days, and carries the highest share of negative sentiment of any platform at 13%. If I were Pizza Hut that's the first place I'd look, and the citations further down show exactly why.
Sentiment by topic
Scroll further down and you get sentiment by topic, based on the topics you set up originally, so you can see which ones are strongest on positive sentiment and which lean negative.
For Pizza Hut, Occasion leads on 84 and Discovery follows on 81. Quality deserves a closer look, since it carries the highest negative share of any topic at 15%. Ordering sits at the bottom on 55, but read the split before reading too much into it. It's 76% neutral and only 7% negative, so AI isn't criticising how you order from Pizza Hut. It just isn't saying much that's positive, which is a very different problem to fix.
We also flag low samples. Where a topic has only a handful of brand mentions across a large number of answers, we won't call it positive, neutral or negative until we've captured more data.
Which publishers are shaping the answer
From there you can drill into inline citations and check which are positive and which are negative. For each domain you can see which platforms it shows up in, how many prompts it covers and the publisher's overall sentiment towards the brand, and competitor sites can be shown or hidden as well.
Over the last 30 days Pizza Hut picked up 2,160 inline citations, with google.com, reddit.com and moneysavingexpert.com the most cited. Most of the top ten score 4 out of 5, but Just Eat comes in at 3, alongside two competitors in Uber Eats and Deliveroo.
Open a domain and you'll see every answer that cites it, split into positive, neutral and negative statements, each with the score we've assigned. Just Eat is cited in 17 prompts, and on the positive side ChatGPT describes Pizza Hut as "a convenient nearby chain option with classic pizzas, sides and meal deals".
The negative side is where it gets interesting. A single Reddit thread comparing Domino's, Pizza Hut and Papa John's is cited in four prompts, and all three answers it feeds are negative. The two shown here both come from Google AI Mode, with one telling users that Pizza Hut's stuffed crust on delivery "can sometimes feel heavier, denser, and less melty" than its rivals. That's the same platform sitting 11 points down on the month, and now you know one of the sources behind it.
This is where it gets practical. You can see whether PR has covered you well, whether your listings and PDPs on retailer sites are accurate and give a fair assessment of the brand, and whether any affiliate coverage positions you in less than the best light. Then you can go and work with those partners to improve it.
The attributes driving the score
We also break down the attributes driving the score, split into what AI praises and what it criticises.
For Pizza Hut, 89% of 25,614 attribute mentions are praise. Price is the clearest example of why the split matters, since it tops both lists: praised 4,991 times and criticised 803 times. Quality and delivery are the next most criticised, which lines up with the Reddit comments above. Further down you'll find cross-contamination, small in volume at 73 mentions but exactly the sort of detail you'd want in front of the right team.
Click into any attribute and you'll see every mention behind it, plus the citations feeding it. Deal has 1,211 mentions, 86% of them positive with an average score of 4.6, including Copilot telling users that ordering direct from Pizza Hut is usually cheaper than going through Just Eat.
Down to the individual prompt
Tracked prompts is where you get the most granular view of sentiment for the answers we've captured. Each prompt shows its score out of 5, the score on every platform, a 30-day trend and the number of citations behind it.
Click into any prompt and you can switch between platforms, follow the capture history day by day and read the answer itself, with every sentence classified as positive, neutral or negative.
Take "Domino's vs Pizza Hut vs Papa John's: which is best for delivery?" in Google AI Overviews. Pizza Hut was mentioned on 28 of the last 29 days, and the 18 October answer scores 67, with three positive sentences, two neutral and one negative. The stuffed crust gets the praise, while the negative flags that delivery times "can occasionally lag during peak weekend hours".
Tell us what you think
We're really pleased with this one and want to hear what you make of it, so drop us an email with your feedback at andy@obsero.ai. If you'd like to see it running against your own brand, get in touch and we'll set up a discovery call.
Frequently asked questions
Is Sentiment Analysis Plus available on every plan?
Yes, it's live for all Obsero customers.
Does Obsero score answers that don't mention my brand?
No. Sentiment is only assessed where your brand is mentioned in the answer.
What does a low sample flag mean?
It means there are only a handful of brand mentions to go on, so we hold off calling that area positive, neutral or negative until we've captured more data.
Growth through intelligence
Capture your brand’s historic visibility, recommendation and sentiment data from AI Search in one intuitive platform