To measure your brand’s AI visibility, write a fixed set of 20 to 50 prompts that your buyers would actually ask, run each one in ChatGPT, Perplexity, Gemini and Google AI Mode on a set schedule, and record five things for every answer: whether your brand is named, where it sits in the list, which sources the answer cites, which competitors appear, and how you are described. Repeat monthly, average the results, and watch the trend rather than any single answer.
You can do it free in a spreadsheet or pay a tool to run it. The hard part is discipline: the same prompts, engines and scoring, month after month.
AI visibility is a trend metric: fix the prompt set, score five things per answer, run it monthly across engines, and judge the direction of the averages, not any single response.
What should you measure in an AI answer?
Record five fields for every prompt, on every engine, every run.
- Named. Does the answer mention your brand at all? Averaged across the prompt set, this is your presence rate, the metric most tools call share of voice or share of answers.
- Position. If you are named, are you first, second, third, or further down? The first two names in a list get most of the attention.
- Sources cited. Which URLs does the answer link to or draw on? This tells you which third-party pages are doing the work. Profound’s analysis of 11.8 billion AI citation data points (Profound, 2026) found that brand websites account for 57 percent of AI citations globally, which leaves about 43 percent to third-party pages such as media, communities and institutions. On ChatGPT, brand websites were 47 percent, so over half of its citations were third-party (WARC summary). If your competitors are being cited from pages you have never appeared on, that is your map.
- Competitors named. List every other brand in the answer. Over time this shows who you are losing to, and on which prompt categories.
- Sentiment and framing. Is the description positive, neutral or negative, and what is the qualifier? “Good for small teams” and “expensive but powerful” send buyers in different directions.
A sixth optional field is accuracy: does the answer get your pricing, features or location right? Wrong facts usually trace back to an outdated third-party page.
How do you build a prompt set worth tracking?
The prompt set is the asset. Change it every month and you have no trend; keep it stable and every run adds a data point. Build it from the questions buyers ask on the way to a decision, not from the keywords you rank for. Aim for 20 prompts at a minimum and 50 if you cover several products or markets. Cover each category below and write prompts the way a real person types.
| Prompt category | What it tests | Example prompts |
|---|---|---|
| Category recommendation | Whether you are in the consideration set at all | “What are the best accounting tools for a small manufacturing business?” |
| Comparison | How you are framed against named rivals | “Compare [Brand] and [Competitor] for a 50-person fintech team” |
| Problem-led | Whether you appear when the buyer has not named a category | “How do I cut invoice processing time without hiring?” |
| Brand check | What the engine believes about you | “Is [Brand] any good? What are its drawbacks?” |
| Local or market-specific | Visibility in the markets you sell to | “Best polymer suppliers in western India for packaging film” |
| Alternatives | Whether you are offered as a substitute | “What are good alternatives to [Market leader]?” |
| Buying detail | Accuracy of facts about you | “How much does [Brand] cost and is there a free plan?” |
Two rules help. First, include at least one phrasing variant for your most important prompts, because a BrightEdge analysis of tens of thousands of identical prompts found that ChatGPT, Google AI Overviews and Google AI Mode disagreed on which brands to name for 61.9 percent of queries (BrightEdge), and in our experience phrasing can shift the list within a single engine too. Second, tag each prompt by category and product so you can read results by segment.
How often should you run the prompt set?
Monthly is the baseline. Weekly during a campaign, a launch or a period of active link building and PR. Daily is rarely worth it, because answers to the same prompt change from run to run for reasons unrelated to you: model randomness, the pages retrieved that day, and wording.
That variation is why you should run each prompt three times per engine and record the average. If a brand appears in two runs out of three, score it 0.67, not 1. A change of a few points month over month is probably noise. As a rule of thumb, look for movement of ten points or more, sustained across two consecutive months, before you call it a trend.
Can you measure AI visibility with a spreadsheet?
Yes, and for most brands this is the right place to start. Set up one sheet with a row per prompt, per engine, per run, and these columns: date, engine, prompt, prompt category, brand named (1 or 0), position (1 to 10, blank if absent), competitors named (comma separated), sources cited (URLs), sentiment (positive, neutral, negative), notes.
Then build a summary tab with four formulas:
- Presence rate = sum of “brand named” divided by number of rows. Calculate it overall, per engine and per prompt category.
- Average position = average of the position column, ignoring blanks.
- Competitor presence = the same presence rate, computed for each competitor you track. Side by side, that is your share of voice table.
- Source frequency = a count of how many times each domain appears in the sources column. Sort it. The top ten domains are the pages that decide your category’s answers.
A 30-prompt set across four engines, run three times, is 360 answers a month: a few hours of copy-paste for one person, or an afternoon with a script that calls the engines’ APIs.
Which tools track AI visibility, and what do they cost?
A paid tool does the running and scoring for you, usually with daily sampling, competitor dashboards and source analysis. Prices change often, so treat the figures below as a snapshot at the time of writing and check each vendor’s page before buying.
| Tool | Approximate list price | Notes |
|---|---|---|
| Otterly.ai | From about $29 a month | Entry plan covers around 15 prompts; cheapest route to automated tracking |
| Peec AI | From about $95 a month (priced in euros) | Starter plan with around 50 prompts and three engines; competitor and source views |
| Semrush AI Toolkit | From about $99 a month per domain | Sold alongside the main Semrush suite |
| Ahrefs Brand Radar | From about $199 a month per AI index, on top of an Ahrefs plan | Strong for seeing which pages are cited across a whole category |
| Profound | Self-serve plans from about $99 a month; enterprise on request | Large citation dataset; built for enterprise teams |
Two cautions. Check which engines the entry plan actually includes, because marketing pages often list engines that only higher tiers cover. And compare tools on the same prompt set, because each vendor computes share of voice differently (some divide by prompts, some by total mentions, some weight by position), so numbers from two tools are not comparable.
How do you read month-over-month movement?
Wait for three months of data before drawing conclusions. Then read the summary tab in this order.
Presence rate first. Is the overall line up, flat or down? Then split it by engine. Perplexity and Google AI Mode retrieve live pages, so they move faster when new third-party coverage appears. ChatGPT without browsing leans more on training data, so it moves more slowly and can lag a campaign by weeks. For how that shapes strategy, see our guide to AEO and GEO (answer engine and generative engine optimization, the work of getting cited in AI answers).
Then source frequency. When presence rises, look at which new domains entered the sources column. If a trade publication, a “best of” listicle or a Reddit thread that now mentions you is being cited, you have found the cause. In engagements we track, brands we have placed in third-party articles, guest posts and listicles have seen AI citations rise by roughly 70 to 80 percent on the prompts where those articles are cited. The source column lets you verify that kind of claim for yourself.
Then competitors. A competitor whose presence jumped usually earned new coverage. Find the pages and work out whether you can be on them too, as explained in why listicles drive AI citations.
Finally sentiment and accuracy. A flat presence rate with improving framing is progress. A rising presence rate with wrong pricing in every answer is a problem to fix on the pages the engines cite.
Report it on one page: presence rate by engine, average position, top ten cited domains, a competitor share of voice table, and three sentences on what changed and why.
How Model Marketing helps
Start with the free LLM Visibility Checker for a quick first read on how visible your brand is to AI engines, then build the prompt set above to see exactly where you are named, where you are missing, and which pages the engines cite instead.
If the picture needs work, our AEO/GEO service begins with a two-week audit: a prompt set of 50 to 200 prompts for your market, a baseline across ChatGPT, Perplexity, Gemini, Claude and Google AI Mode, a source map of the third-party pages that decide your category’s answers, and a review of how quotable your own pages are. We scope the ongoing program after the audit, month to month with no retainer; details are on the pricing page.
The work that moves the numbers is mostly off-site, which is why we run AEO alongside link building, digital PR and community marketing. Engines recommend the brands the rest of the web vouches for, whether that brand sells software, polymers, financial services or consumer goods. Measurement tells you which pages to earn a place on.