Reading time: 17 minutes
GEO tools measure how your brand shows up in answers from ChatGPT, Gemini and Google’s AI. In general, each GEO tool runs lists of prompts, stores the answers and counts mentions, citations and competitors. This comparison applies six measurement criteria to seven tools. So that you can check every figure, we only recorded what the official documentation said.
In short
- Ahrefs, Profound and Otterly document collection through the engines’ web interface, as the user sees it. On the other hand, the two Brazilian tools in this comparison don’t state their method.
- Five GEO tools separate brand mentions from site citations in AI answers. However, none of the seven separates recommendations from mentions.
- Four GEO tools list Brazil among their countries: Semrush, Otterly, Ahrefs and Peec AI. In addition, four document exporting the full text of the answers.
- Entry plans range from $29 to $99 per month, at the prices checked for this guide. At Profound, however, one agency client costs $498. Radar GEO is free, while GEOscore doesn’t publish a price.
- Outside the comparison, SEONextbr, Promptado, First Answer and naia publish prices in reais, starting at R$97 a month.
What is a GEO tool?
A GEO tool runs a fixed list of prompts on AI engines, stores the answers and counts how many of them include the brand. It also shows which sites the AI cites and which competitors show up alongside it. It automates the collection the team would otherwise do by hand, with the date, engine and country recorded.
GEO is the work of getting a brand into those answers, as our complete guide to GEO explains. So GEO tools measure the effect of that work on what the AI answers. For WordPress sites, Yoast, Rank Math and AIOSEO also offer AI measurement in paid plans; see GEO plugins for WordPress.
This comparison covers five international GEO tools: Semrush AI Visibility Toolkit, Ahrefs Brand Radar, Profound, Peec AI and Otterly. It also covers two Brazilian tools: Radar GEO, from RD Station, and GEOscore, from MonitorCNPJ. NotHuman, which publishes this blog, also offers monitoring.
Which criteria should you use to choose a GEO tool?
Use the four vendor questions from the measurement protocol with fixed, repeated prompts we published on this blog. Here, however, they become six criteria, with the engines covered listed first. For each criterion, we recorded only what the official page or documentation of each GEO tool said. So anything the page doesn’t say appears as “not stated”.
- Engines covered: which AIs the plan includes.
- Interface or API: whether collection reproduces what the user sees.
- Country and language: where the tool sends the prompt from and in which language.
- Runs per prompt: how many times each question runs.
- Mention, citation and recommendation: whether the three metrics come separately.
- Answer export: whether the full text comes out for auditing.
Finally, price comes separately, as published and in the original currency.
Which engines does each GEO tool cover?
All seven GEO tools cover ChatGPT, and the five international ones include at least one Google AI feature. However, the difference is in the plan: Gemini, AI Mode and Claude are often add-ons or sit in the bigger plans. So check the engines in the plan you’ll buy, not just on the features page.
| Tool | Engines in the documentation |
|---|---|
| Semrush | ChatGPT, AI Mode, Gemini and AI Overviews |
| Ahrefs Brand Radar | AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot; Claude only in custom prompts |
| Profound | ChatGPT, Gemini and AI Overviews in the trial; up to 9 in Enterprise |
| Peec AI | 3 of your choice in Starter; up to 13 in Enterprise |
| Otterly | ChatGPT, AI Overviews, Perplexity and Copilot; AI Mode, Gemini and Claude as add-ons |
| Radar GEO | ChatGPT and Gemini |
| GEOscore | ChatGPT, Gemini, Copilot, Perplexity, Claude and others on request |
Ahrefs notes that it temporarily can’t collect new Grok data because of a policy change by that engine. At Profound, a client on the agency plan tracks ChatGPT, Perplexity and AI Overviews. The other engines, though, sit in Agency Enterprise.
Does the tool collect through the interface or the API?
Ahrefs, Profound and Otterly say they collect through the engines’ web interface, as the user sees it. Semrush, however, says the same only for the prompt database behind its reports. In contrast, Peec AI’s pricing page flags the models it runs through the API, such as Claude, Grok and DeepSeek. Radar GEO and GEOscore don’t say how they collect.
This matters because the API can cite different sources from the ones a consumer sees in the app. So the protocol recommends collecting through the interface.
According to Ahrefs, Brand Radar types the questions into the web version of each chatbot. It uses the default model, with no history or user data. But Claude is the exception, and Ahrefs tracks it through the API. Likewise, Otterly posts the prompts on each engine’s website and uses the API only for Claude.
In the same way, Profound says it captures answers directly from the consumer browsing experience, not through the model APIs. Semrush says its prompt database comes from real requests, with no API. For Prompt Tracking, however, it only says it queries the platform in the chosen location.
Can you collect from Brazil and in Portuguese?
Yes. Semrush and Otterly list Brazil in their help centers. In addition, Ahrefs and Peec AI include the BR code in the country list of their prompt API. Profound doesn’t publish its list of regions, but its prompt setup guide uses Brazil and Portuguese as an example.
Radar GEO and GEOscore, however, don’t state the country. So ask the vendor to confirm in writing that the plan includes Brazil and Portuguese.
| Tool | Country and language |
|---|---|
| Semrush | More than 220 countries in Prompt Tracking; Brazil on the list for AI analyses |
| Ahrefs Brand Radar | Country per prompt; BR on the API list; language not stated |
| Profound | More than 30 languages and 150 regions; Brazil and Portuguese cited as an example |
| Peec AI | Country per prompt, with a limit per plan; BR on the API list; languages at no extra cost |
| Otterly | More than 70 countries; Brazil in all 7 engines; any language |
| Radar GEO | The company’s segment and location; language not stated |
| GEOscore | Not stated |
Otterly says it queries the engines as if a real user were searching from inside each country. The prompt also goes out exactly as written, in any language. At Ahrefs, on the other hand, each extra location uses one more check. Semrush flags a limit: Prompt Tracking only simulates desktop searches.
How many times does each prompt run?
Semrush, Profound, Peec AI and Otterly run the prompts every day. Ahrefs lets you choose daily, weekly or monthly, while GEOscore updates every week. Profound is the most explicit: each prompt runs once a day on each engine. However, no documentation mentions repeating a prompt within the same cycle.
One run a day adds up to about 30 answers per prompt and engine each month, more than the protocol’s 5 runs. Repeating matters because the answer changes with every run. Rand Fishkin, cofounder of SparkToro, changed his mind after measuring this variation across 2,961 runs:
“I’ve been swayed from my initial position and now believe visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric.”
Rand Fishkin, SparkToro (2026)
Patrick O’Donnell, of AI tracking startup Gumshoe.ai, took part in the study. So Fishkin himself points out the possible conflict of interest.
In the same piece, he also criticizes tools that give a ranking position in AI. Still, Peec AI, Profound and Otterly show the brand’s average position in the answer, and GEOscore shows its position against competitors. Semrush shows the domain’s average position among the cited sources. Therefore, use these numbers only as context, not as a client goal.
Does the tool separate mentions, citations and recommendations?
Five of the seven tools separate brand mentions from site citations, each with its own metric name. However, none of the seven documents a recommendation rate separate from mentions. The closest is Peec AI’s win rate, which counts the answers where the brand comes first.
| Tool | Mention and citation | Recommendation |
|---|---|---|
| Semrush | Separate | Not stated |
| Ahrefs Brand Radar | Separate | Not stated |
| Profound | Separate | Not stated |
| Peec AI | Separate | Not stated, but has win rate |
| Otterly | Separate | Not stated |
| Radar GEO | Mention, but cited site not stated | Not stated |
| GEOscore | Presence, but cited site not stated | Not stated |
Ahrefs and Peec AI go one step further. Both separate the page cited in the text from the page the engine only consulted. That list of sources is the raw material for an analysis of the sources AI cites about your brand.
To measure recommendations, the protocol suggests flagging answers that present the brand as an option. For example, the answer may list the brand or use a verb like “consider”. With the exported answers, the team then calculates that rate on its own.
Can you export the full answers?
Four GEO tools document exporting the text of the answers: Otterly, Peec AI, Profound and Ahrefs, the last one through the API. After all, export lets you audit the metric, recount mentions and classify recommendations and sentiment with your own criteria. So without it, nobody can check the number a GEO tool shows on its dashboard.
| Tool | Documented export |
|---|---|
| Semrush | Excel, CSV, PDF and Google Sheets, but full text not stated |
| Ahrefs Brand Radar | Answer text through the API, as long as the plan has API access |
| Profound | Answers in CSV and JSON in Enterprise and in agency client workspaces |
| Peec AI | All chats, from the project tab |
| Otterly | Raw answers in CSV and JSON, on all plans |
| Radar GEO | Not stated |
| GEOscore | Monthly PDF report, but answers not stated |
At Otterly, raw answers export as CSV or JSON on all plans. Also, at Ahrefs, pulling custom prompt data through the API doesn’t use up units. Semrush shows the original answer in a pop-up window, but doesn’t say whether the full text goes into the export. Finally, at Profound, the free trial exports nothing. Before you resell the measurement under the agency’s brand, see what to check in our guide to white label GEO.
How much do GEO tools cost?
Entry plans range from $29 to $99 per month, at the prices checked for this guide. For most GEO tools, the price grows with the number of prompts and AI engines. In addition, at Ahrefs and on Peec AI’s agency plans, the price grows with collection frequency.
Profound, on the other hand, charges per client on its agency plan: one client comes to $498 per month. Agency plans, client access and white label are in the guide to AI visibility tools for agencies.
| Tool | Published price | What’s included |
|---|---|---|
| Semrush | $99/month | 25 prompts, daily |
| Ahrefs Brand Radar | $50/month (Basic) | 2,500 checks |
| Profound | $99/month (Agency Growth) | 7-day audits; each client, +$399/month |
| Peec AI | $95/month (Starter) | 50 prompts, 3 engines |
| Otterly | $29/month (Lite) | 15 prompts, 4 engines |
| Radar GEO | Free | Diagnosis in ChatGPT and Gemini |
| GEOscore | Not published | Quote on request |
The figures come from Semrush’s documentation and the pricing pages of Ahrefs, Profound, Peec AI and Otterly, on the monthly plan. Otterly says its prices don’t include taxes.
Some add-ons, however, change the math. At Otterly, for example, Gemini and AI Mode cost $9 per month each on Lite, and Claude costs $29. Ahrefs includes 5 prompts in its Lite plan, at $129. Brand Radar AI, with Ahrefs’ own prompt database, starts at $199.
Profound has a 7-day free trial for brands and an Enterprise plan with pricing on request. Otterly and Peec AI also offer a free trial. Semrush, however, doesn’t offer a trial of its AI toolkit.
Which Brazilian GEO tools exist?
At least eight Brazilian GEO tools track brands in AI answers and have an official page. But only two appear in this comparison’s tables: Radar GEO, from RD Station, and GEOscore, from MonitorCNPJ. Radar GEO is free and runs a diagnosis in ChatGPT and Gemini. GEOscore, on the other hand, works on a quote basis, with weekly updates across eight AI platforms.
Radar GEO shows brand mentions within the segment and location, sentiment and a comparison with competitors. There’s also a version of the page for agencies. However, the page doesn’t say how many prompts the tool runs, how it collects or whether it exports the answers.
GEOscore combines presence in AI, Google and social media into an index from 0 to 100. The page mentions a monthly PDF report and an action plan, but doesn’t state the collection country or method.
How much do the other Brazilian GEO tools cost?
Of the other six, four publish prices in reais, with entry plans from R$97 to R$1,490 a month. Geo Pulse and TEIA GEO IGO, on the other hand, work on a quote basis.
| Tool | What it tracks and on which engines | Published price |
|---|---|---|
| SEONextbr | Brand visibility in ChatGPT, Gemini and other AIs, within an SEO suite | R$97 a month, with 10 AI visibility queries |
| Promptado | Mentions, citations and main recommendation in ChatGPT and Gemini | R$399 a month, with 20 prompts |
| First Answer | Mentions and cited sources in ChatGPT and Gemini | R$690 a month, with 50 prompts |
| naia | GEO score in ChatGPT and Gemini, plus AI Overviews | R$1,490 a month, with 4 analyses a month |
| Geo Pulse, from Spark Group | 11 metrics, such as share of recommendation, in ChatGPT, Gemini, Claude and Perplexity | On request |
| TEIA GEO IGO | How ChatGPT, Gemini, Claude and Perplexity interpret the brand | On request |
The full comparison is in our guide to the best AEO and GEO tools in Brazil for each use case.
When is it worth building your own collection?
Building your own collection makes sense in two cases. The first is when none of the tools documents something essential to you. The second is when the price per client times your client base exceeds the cost of a script. The math starts from the protocol in our measurement article: 1,000 answers per month per client.
In other words, that’s 50 prompts, 5 runs and 4 engines. The table uses 21 business days per month.
| Client base | Answers per month | Per business day |
|---|---|---|
| 1 client | 1,000 | about 48 |
| 5 clients | 5,000 | about 238 |
| 15 clients | 15,000 | about 714 |
So at this volume, manual collection stops being feasible from the very first client.
How much does the protocol cost in each GEO tool?
To compare, translate the protocol into each tool’s billing unit:
- Ahrefs: one check equals one run of one prompt on one engine and in one location. So the protocol equals 1,000 checks per client per month, and weekly frequency comes close to the 5 runs. The Scale package, at $250, includes 25,000 checks and covers 15 clients without Claude. That’s because each Claude query uses eight checks.
- Profound: each client costs $399 per month, on top of the $99 for Agency Growth, with 100 prompts and three engines. As a result, the total for 15 clients comes to $6,084 per month.
- Otterly: Standard, at $189, has 100 prompts on four engines, or 150 for agency partners. That fits the lists of two or three clients. On this plan, however, Gemini and AI Mode cost $59 per month each.
- Peec AI: Starter, at $95, covers one client’s list on three engines. For agencies, on the other hand, Essential, at $245, has 111 daily prompts on three engines, which fits two clients.
- Semrush: the plan includes 25 prompts, but one client’s list needs 50. Each additional 50 prompts costs $60 per month.
On the script side, in contrast, the cost is development, infrastructure and maintenance every time the engines’ interface changes. Then compare that cost with the price per client times your client base.
A script also covers what no documentation describes, such as several runs of a prompt on the same day. Collection cost per client is the first line in the math of how to build a GEO service at your agency.
What we found when we analyzed seven GEO tools
We read the pricing pages, documentation and help centers of the seven tools. Then we applied the six criteria and counted how many tools meet each one. Overall, the biggest gaps are in recommendations, repeating a prompt within the same cycle and exporting the answers.
- Engines: 7 of 7 state them.
- Interface or API: 5 of 7 state it, but Semrush only for its prompt database.
- Brazil: 4 of 7 list the country, Semrush and Otterly in their help centers, Ahrefs and Peec AI in their API documentation; Profound cites Brazil as an example.
- Frequency: 6 of 7 state it, but none mentions more than one run per cycle.
- Mention and citation: 5 of 7 separate them, while separate recommendation stands at 0 of 7.
- Export of answer text: 4 of 7.
- Price: 5 of 7 publish a paid plan, while Radar GEO is free and GEOscore works on a quote basis.
Among Brazilian GEO tools outside the tables, however, Promptado separates a main recommendation from a passing mention in AI answers.
GEO tools measure what the AI answers, but they don’t change the answer. If the brand doesn’t show up yet, the earlier step is to understand how to show up in ChatGPT and the other engines.
Frequently asked questions
Yes. Radar GEO, from RD Station, is a free tool and runs a brand diagnosis in ChatGPT and Gemini. The Brazilian GEO tool naia has a free plan, with 2 prompts in ChatGPT and AI Overviews. Profound, Otterly and Peec AI also offer free trials, and Profound’s trial for brands lasts 7 days. Semrush, however, doesn’t offer a trial of its AI toolkit.
No. A GEO tool measures what the AI says about the brand, but it doesn’t change the answer. So that the brand shows up, the work is technical access, indexing and citable content. Our guide on how to show up in ChatGPT covers this work.
Average position is useful only as context, not as a client goal. Because the AI’s answer changes with every run, Rand Fishkin, of SparkToro, criticizes tools that give a ranking position in AI. Instead, he sees the visibility percentage across many prompts, repeated several times, as a reasonable metric.
Want to know how your brand shows up in AI answers today?
NotHuman is opening spots for agencies and brands to track how they appear in AI answers. Join the waitlist and we’ll let you know when a spot opens up for you.
Main sources
- Semrush, where the AI Visibility Toolkit data comes from (2026).
- Ahrefs, what Brand Radar is and how it collects answers (2026).
- Profound, plans for brands and agencies (2026).
- Otterly, how the tool collects data (2026).
- RD Station, Radar GEO, page in Portuguese (2026).
- SparkToro, study on how inconsistent AI brand recommendations are (2026).
