Brand Search: Tracking Direct Brand Research
Centium now tracks what AI models say when prospects research your brand directly, alongside the category recommendations we have always measured.
the other half of the picture.
Centium's core methodology prompts AI models at the category level. We ask the questions real buyers ask, like "best activewear brands for running," and measure which brands the models recommend. Prompting at the category level instead of naming the brand is what keeps the data unbiased.
Brand search adds the other half. It measures what ChatGPT, Claude, Gemini, Perplexity, and Grok say when a prospect researches your brand directly. Questions like "is this brand worth it" or "what do people say about this brand" produce a different kind of answer, and now you can see both.
Here is what that looks like in practice, with real data from the demo dashboard. Same brand, same five models, two different questions. Toggle between Category and Brand.
We tend to see category searches go heavily toward media outlets, ecommerce platforms, and sites that aggregate data about your industry. Brand search skews to your own website, Reddit threads about you, and the places AI finds when investigating your brand.
recommendation vs. direct research.
AI talks about your brand differently depending on the question. When a model recommends you inside a category answer, the language tilts positive. The model has already decided you belong on the list, and it describes you in the terms that justified the pick.
When a prospect researches you directly, the same models weigh the full record. They surface pricing concerns, quality complaints, shipping issues, and competitor comparisons that the recommendation context never showed. This is the conversation your buyers are actually having in the final stretch before a purchase, and it was previously invisible.
The most useful read is the sentiment shift between the two views. A theme that sits Positive in the category view and Mixed or Negative in the brand view is friction your prospects are hearing at the moment of decision. That gap is where the work is.
| Theme | Frequency |
|---|---|
Exceptionally Soft Fabrics | |
Versatility (Gym-to-Street) | |
Modern Minimalist Design | |
Technical Performance Fabrics | |
Comfort-Focused Design | |
Sustainability & Eco-Conscious | |
Rapid Growth & Rising Popularity | |
Men's & Unisex Focus | |
Durability & Quality | |
Travel-Friendly | |
Popular Core Products (Joggers/Shorts) | |
Premium Price Point |
where brand search lives.
Brand search runs a dedicated set of brand-level prompts through the same five models and the same processing pipeline as your category prompts, extracting the themes AI uses to describe you, the sources it draws on, and the citations it points users to. Sections that carry both views show a Category | Brand toggle in the card header:
Themes and sentiment in each context, side by side. Tagged Positive, Neutral, Mixed, or Negative.
The prompts that drove citations to your domain, split by context.
The sources AI leans on for each question type. Expect far more citations about your brand.
Where models point users for more about your brand, including social channels.
A prompt that names your brand surfaces you nearly every time. Counting it would pin your rate at 100%.
Direct brand questions are not a competitive context. Including them would distort the comparison.
Brand search exists for perception and sourcing: what AI tells customers who ask about you, and which sources it pulls from when it answers.
same methodology, fewer limits.
Brand search does not change your category data. Category prompts remain the unbiased core of the methodology, and your visibility trends stay fully comparable across runs. Brand-level prompts are a separate set, processed and reported separately, so each view answers its own question: the category view tells you whether AI recommends you, and the brand view tells you what AI says when someone checks.
It ships alongside a second change worth knowing about: the six category cap is gone. Your 120 prompt allotment is yours to distribute across as many categories as you like, and every prompt runs across all five models. That makes room for a brand category without giving up market visibility coverage, and it means you can weight your prompt mix however your strategy calls for.
turning it on.
How you turn it on depends on where your brand is: new brands get it during onboarding, existing brands add it from the prompt editor.
New brands
If you are setting up Centium for the first time, brand search is part of the walkthrough. Step 2 of onboarding suggests your measurement categories: six for market visibility and one brand category, pre-named "About [your brand]" and already set to Brand.
Here is that step live with demo data. The toggles, remove buttons, and Add Category all work.
Add or remove categories freely within your 120 prompt allotment, and mark any row Brand or Non-Brand with the toggle.
MEASUREMENT CATEGORIES
Confirm your industry and choose your measurement categories. Centium recommends 7: six for market visibility and one for brand perception. We'll generate 120 prompts across them.
Existing brands
If your brand is already running on Centium, open your dashboard and click Edit Prompts in the menu. In the editor, click Add next to Categories and create a new category with the Brand type. This is the dialog you will see, live with demo data.
Add a new brand category rather than switching an existing category to Brand. Your existing categories keep measuring the same question they always have, so their trend lines stay comparable, and the new brand category starts a clean baseline of its own.
The exception: if you already use a category to research your own brand terms, switch it to Brand. Those prompts were measuring direct brand research all along, and assigning the category correctly separates those insights from your market categories so they no longer distort your visibility rate or your sourcing data.
Adding a category will likely push your prompt count past 120. That is expected. Add everything you want, then remove prompts from any category until the counter at the top of the editor is back at 120. You decide what gets trimmed to make it fit.
Add New Category
Enter a category name and we'll generate prompts for it.
19/40 characters
Saved changes apply on your next scheduled update.
questions, answered.
Yes. Brand categories draw from the same 120 prompt allotment as your market categories, and you decide the split. Every prompt runs across all five models either way.
For new brands, Centium distributes them automatically across the categories you select. The default seven categories (six non-brand, one brand) builds 17 prompts per category. Choose eight and Centium builds 15 per. Existing brands adding a category will likely go past 120, so expect to eliminate some prompts from other categories to make room.
No. Brand search categories are excluded from Visibility and Competitors. A prompt that asks about your brand by name surfaces your brand nearly every time, so counting it would pin your visibility rate at 100% and distort competitor comparisons.
Brand Perception, Referral Drivers, Websites Referenced, and Citations all carry a Category | Brand toggle. In Brand Perception, the toggle splits how AI describes you when recommending you in market answers versus when prospects research you directly.
In the websites and sourcing sections, it splits where AI pulled its information from: the category view shows the sources behind the broader market search terms, and the brand view shows the sources behind questions about your brand.
Add a new one in most cases, so your existing categories keep their trend history. The exception is a category you already use to research your own brand terms. Switch that one to Brand so those insights are typed correctly and stop distorting your visibility and sourcing data.
New brands get it during onboarding, where Centium recommends one brand category by default. Existing brands open their dashboard, click Edit Prompts in the menu, then add a category with the Brand type.
Saved prompt changes apply on your next scheduled update.
More guides
The Order Your Brand Appears in AI Answers
Visibility tells you whether AI names your brand. Position Frequency shows where you land in the answer next to competitors, across ChatGPT, Claude, Gemini, Perplexity, and Grok.
Now Tracking Claude Sonnet 5
Centium now tracks Claude Sonnet 5. Its reliable knowledge cutoff moved from August 2025 to January 2026, putting more recent information behind the answers we measure.
How PR and Earned Media Drive AI Visibility
PR used to be a long game you could not measure. Now earned media is a performance channel, and Centium ties your coverage to AI citations and visibility.