Getting Started With Centium
What Centium does, how the measurement works, and what happens after you sign up. Start by building the segments we would measure for your brand.
ai is how people find brands now.
ChatGPT passed a billion monthly active users in June 2026, and Google reported the Gemini app at 950 million monthly actives in July. Both figures sit before you count AI answers inside Google search itself. Two years ago none of this existed at scale.
What makes it worth measuring is the quality of the traffic. Adobe found that visitors who arrive at a website from an AI model buy 42% more often than visitors from any other source, and you can track that traffic against your own brand. A model recommendation carries conviction in a way a page of links never did. It works like an influencer merged with a search engine: it has opinions, it makes a pick, and people act on it.
So the practical question for a brand is no longer whether to pay attention. It is how to find out where you stand today, and what to do about it. That is the whole job of this guide.
we give AI an exam.
AI models report almost nothing back. There is no Search Console for ChatGPT, no keyword report, no impression count. Part of that is privacy and part of it is that the question is genuinely hard: a recommendation often arrives twenty prompts into a conversation, so there is no clean query to log.
The way around it is to ask. Centium prompts the models hundreds of times about your categories on a schedule, records every answer, and reports what came back. Think of it as a standing focus group of AI models, or an exam you give them once a month. Two rules keep the results honest:
- Most prompts never mention you. They ask what your buyers ask, like “best insulated water bottle for backpacking.” If a model brings you up on its own, that counts. We do point a smaller set of prompts straight at your brand, because that is what a prospect hears when they research you, but those answers are reported on their own and kept out of your recommendation rate.
- Ask many times, many ways. AI is variable by design. The same question can return different brands on different runs, so one answer is an anecdote. You flip the coin a hundred times, not ten, and the pattern that survives is the real one.
Both rules point to the same conclusion. You cannot measure this one keyword at a time. You measure it by category, which is where segments come in.
the categories you measure.
AI has no ranked results page. There is no position one through ten to climb and no bid to place, so there is nothing to track keyword by keyword. Segments are how you decide what to measure instead.
A segment is one slice of your market that you want prompts pointed at. Split them by product line, by geographic market, by customer type, or however your strategy is organized. An activewear brand might measure athleisure, trail running gear, and compression. A resort might measure family travel, summer, and one segment per feeder market.
Each segment is scored on its own. In this demo dashboard the brand is named in 90% of athleisure answers and 13% of clothing answers, on the same run.
Filter by model and it splits further, since a segment can be strong on Gemini and invisible on ChatGPT.
120 prompts, six models.
Once your segments are set, Centium writes the prompts. Every plan includes 120 unique prompts distributed across your segments, and each prompt runs on all six models: ChatGPT, Claude, Gemini, Perplexity, Grok, and Google AI Mode. That is 720 answers behind every update.
Centium recommends seven segments by default: six pointed at your market to measure visibility, and one pointed at your brand for sourcing and perception. That puts roughly 15 to 20 prompts in each, depending on how many segments you end up adding.
The prompts are editable the same way the segments are. Change a line, delete one, write your own. An AI assistant is built into the editor, so you can steer the whole set in plain language, something like “lean more family focused,” and regenerate.
Here is what they look like once a run finishes. Each prompt shows which models named you, and the dimmed rows are blind spots where none of them did.
That seventh segment is worth understanding before you submit. Models get noticeably more critical when you ask about a brand directly than when they are recommending one, and both readings are useful. Centium keeps them apart so the brand questions never inflate your visibility number. There is a full guide on brand search if you want the detail.
| Prompt | Models |
|---|
see your own segments.
The fastest way to understand any of this is to run it on your own brand. Enter your domain and Centium reads your site and builds the segments it would recommend, with sample prompts under each one and the subtopics it would use to steer them.
Underneath sits a training layer we built from thousands of prompts about how customers actually search, so what comes back is phrased the way buyers ask rather than the way brands write. It is the same engine that runs during onboarding.
Outdoor Backpacks
20 promptsBest outdoor backpacks for beginners
Top rated outdoor backpacks
What outdoor backpacks should I buy
Hiking Apparel
20 promptsBest hiking apparel for beginners
Top rated hiking apparel
What hiking apparel should I buy
Trail Running Gear
20 promptsBest trail running gear for beginners
Top rated trail running gear
What trail running gear should I buy
Camping Equipment
20 promptsBest camping equipment for beginners
Top rated camping equipment
What camping equipment should I buy
Whatever it gives you is a starting point, not a verdict. In onboarding every line is editable: rename a segment, remove one that does not apply, add the one the scan missed, and add subtopics to steer the prompts toward specific products.
measure. understand. optimize.
Most AI monitoring stops at the first step and tells you whether you got mentioned. Your dashboard answers three questions in sequence, and each one is more useful than the last.
Are you visible?
We run 120 prompts across six AI models, 720 answers in all, to reveal how often your brand and competitors appear in the searches your potential customers are making.
Learn how AI thinks.
We treat each AI model like a persona. What sources do they consume? What do they consider important when forming a recommendation? We study how each one thinks so you know exactly what influences their decisions.
Optimize and impact.
We turn insights into a personalized action plan with specific tasks and priorities, so you can position your brand exactly where AI models are looking. Then we track your progress over time to measure success.
Where each answer lives
Your dashboard carries a section for each: visibility and competitors, perception and decision factors, sources and citations, query fan-out, indexing, and attribution. The platform overview walks through all of them.
The plan at the end is built from your own data, so it is specific: pitch the editors at the outlet cited in a third of your category's answers, publish the comparison page the models kept searching for, join the subreddit that keeps turning up in your sources. It changes every update as the data moves.
Worth being straight about the boundary: Centium is intelligence, not execution. It tells you where you stand and what to do. The pitching, writing, and outreach is still yours or your agency's. Most of what works is PR and earned media, and most of the rest is publishing the kind of evidence AI actually reads.
from signup to first dashboard.
Simple steps, with no kickoff call and nothing to install.
- 01
Subscribe online
1 minutePick your cadence on the pricing page and check out. No setup fees, no sales call, no contract to sign, and you can cancel at any time.
See plans - 02
Enter your brand and website
1 minuteA name and a URL is the whole form. You can add background context if you want to steer the scan, but the recommendation engine does well without it.
- 03
Review your segments
1 minuteCentium proposes seven from your site: six market segments and one for your brand. Rename, remove, add, and set subtopics until the list matches how you think about your market.
- 04
Generate and adjust your prompts
5 minutesCentium writes 120 prompts across your segments. Edit any line, or steer the whole set with the built-in AI assistant and regenerate. Save your progress at any time and hit Submit when you are ready.
- 05
Your first dashboard gets built
same dayOur team inspects your segments and prompts, and processes the dashboard once everything looks right: 120 prompts across six models, 720 answers. Typically within one hour, always the same day, and you get an email when it is live.
- 06
Then it refreshes automatically
ongoingThat review happens once. After your first dashboard, updates run on their own at the cadence you picked, with nothing to trigger. You can edit your segments and prompts at any time, and changes go live on your next update.
After you are live
- Prompt edits land on the next run. Open Edit Prompts from your dashboard menu at any time. Your next update date sits on the dashboard, so get changes in before it.
- Add anyone you want. Unlimited users, each set as a viewer or a manager. Managers can edit prompts, viewers can only read the dashboard.
- Or skip accounts entirely. Generate a share link that works like a Google Drive link, with an expiration date if you want one, so a client or a colleague can open the dashboard without signing up.
- Cancel whenever. Your dashboard stays live either way, so the history you have built does not disappear.
From here, how we measure AI visibility goes deeper on the methodology, and how AI answers questions explains why some brands move quickly and others take a season.
questions, answered.
AI visibility is how often AI models name your brand when someone asks for a recommendation in your category. If you ask ChatGPT for the best trail running shoes a hundred times and your brand comes up thirty of them, your AI visibility in that category is 30%. It is the closest thing this space has to a ranking, and it is the number Centium is built to measure and move. You will also hear the practice called AEO, for answer engine optimization.
No. Setup is a brand name and a website URL. Centium reads your site and proposes the categories to measure and the prompts to run, and you edit them in plain language. There is nothing to install, no tracking code, and no data feed to connect. Connecting Google Analytics is optional and only affects the Attribution section.
Most brands finish the setup in about ten minutes. There is no questionnaire and no kickoff call, because Centium generates your starting point from your website instead of asking you to build it. Once you submit, a person on our team reviews your setup before anything runs, and your first dashboard is typically ready within an hour and always the same day. Every update after that is automatic.
A segment is one slice of your market that you point prompts at, like “trail running gear” or “the Denver feeder market.” You choose them, and you can split by product line, geography, customer type, or however your strategy is organized. AI answers have no rankings to track keyword by keyword, so segments are the unit of measurement instead. Each one holds its own prompts and gets its own recommendation rate.
Centium recommends seven by default: six pointed at your market to measure visibility, and one pointed at your own brand for sourcing and perception. That puts roughly 15 to 20 prompts in each. You can add or remove segments freely, and the 120 prompts redistribute across whatever you end up with.
Yes, in one place and on purpose. Your six market segments never name you, because naming your brand biases the answer and would measure nothing. The seventh segment asks the models about you directly, which is what a prospect hears when they research you before buying. Those results are reported separately and excluded from your recommendation rate, so the two never mix.
Six: ChatGPT, Claude, Gemini, Perplexity, Grok, and Google AI Mode. Each one sources information differently, so they are reported separately as well as blended. Every prompt in your plan runs on all six, which turns 120 prompts into 720 answers per update.
Yes, at any time. Open Edit Prompts from your dashboard menu, change whatever you want, and save. Changes apply at your next scheduled update, and the date is shown on your dashboard. One caution: changing prompts mid-campaign resets the comparison, so a segment you plan to trend is worth leaving alone once it is running.
You pick the cadence: monthly, bi-weekly, or weekly. There is also a single analysis product if you want one report with no subscription.
No. Centium is an intelligence platform. It measures where you stand, shows how the models made their decisions, and writes a specific plan, but the pitching, publishing, and community work is yours or your agency’s. That is the honest boundary of the product today.
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.