How AI reads this page
ChatGPT, Claude, Gemini and Perplexity don't look at colours, images or animations. They receive text, headings, links and structured data. Below is exactly what they read when they open “We Dropped the APIs and Now Simulate the Real User”.
The AI receives 1,023 words in 10 sections, one link and 3 blocks of structured data.
Text
The page content in Markdown, without menus or graphics. It is the format closest to how a language model reads text.
# We Dropped the APIs and Now Simulate the Real User > Why Geosnap doesn't measure AI visibility through APIs: it simulates real users on ChatGPT, Gemini and Perplexity to capture the answers customers see. URL: https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale Language: English Version IT: https://geosnap.ai/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale Author: Rinald Sefa, CMO Geosnap · Category: Product Publisher: Geosnap (Maind Group S.r.l.), https://geosnap.ai ## In short Geosnap doesn't query AI models via API, because API calls run in a controlled technical environment that doesn't reflect what a real user sees and can produce partial or misleading data. Instead, it simulates the full interaction on ChatGPT, Gemini and Perplexity, from the browser and session context to the question itself, and repeats the process every month on the same prompts to get data that's comparable over time. - API calls query models in a controlled environment and can give a partial or misleading picture. - Geosnap simulates real users on ChatGPT, Gemini and Perplexity with prompts generated from brand, industry and market. - AI Rating, Sentiment and Sources are based on real answers, collected every month on the same prompts. Translated from the Italian original. [Read the original](https://geosnap.ai/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale) ## How we analyze AI visibility: why we abandoned APIs and simulate the real user ## The problem nobody was tackling When we started working on AI visibility, the first question we asked ourselves was a simple one: how do we know what a user actually sees when they ask ChatGPT something? The answer everyone gave was just as simple: you call the model via API, send the question, read the answer. Fast, scalable, automatable. The problem is that this isn't what happens in reality. ## API vs. real experience: a difference that matters When a potential customer looks for a solution on ChatGPT, they don't use an API. They open the browser, go to the platform, start a conversation. The context in which that conversation takes place, the open session, the mode of interaction, the way the question is phrased, all of this influences the answer the model returns. API calls bypass all of this. They query the model in a technical, controlled environment that doesn't reflect the user's real experience. The result is data that looks precise but tells a partial story, sometimes a misleading one. We've seen brands that, according to API analyses, appeared well positioned, but in practice never showed up in the answers a real user would have received. And vice versa. ## The solution: simulating the real user We decided to tackle the problem at its root. Instead of calling the models via API, we simulate the user's real interaction, step by step. The process is this: we open the browser, access the AI platform, whether it's ChatGPT, Gemini or Perplexity, build the session context exactly as a real user would, enter the question the way a person would write it, and read and analyze the complete answer. It's not a trivial process. It requires significant technical infrastructure, because it has to be repeated for hundreds of prompts a month, across multiple platforms, in a consistent and measurable way. But it's the only way to get data that truly reflects what your potential customer sees. ## Why this changes everything The difference between API data and data based on real user simulation isn't just technical. It's strategic. If you're building an AI visibility strategy on data that doesn't reflect reality, you're optimizing for an environment that doesn't exist. You're improving your position in a lab test while, in real life, your competitors are being recommended instead of you. With real simulation, the data you see in the dashboard is the same data your potential customer would see if they ran that search right now. No technical mediation, no artificial environment. What you measure is what matters. ## How it works in practice Every analysis we run on Geosnap follows this process. Prompts are generated dynamically, based on the client's brand, industry and specific market. They aren't generic questions; they're the questions a real user would ask when looking for a solution like yours. These prompts are then run through the simulation on ChatGPT, Gemini and Perplexity, collecting the answers exactly as a user would see them. The results are analyzed to measure mention frequency, context, sentiment and the sources used. The process is repeated every month, on the same set of prompts, so that the data is comparable over time and progress can be measured meaningfully. ## What you see in the dashboard The data collected through the simulation feeds every section of the platform. The AI Rating you see reflects how often your brand appears in real answers, not in API calls. Sentiment analyzes how it's described in those real answers. Sources shows which sources the models drew on to build those real answers. It's a difference that seems subtle, but it changes the meaning of every data point you read. ## The result for Geosnap users Companies that work with Geosnap make decisions based on what actually happens in their potential customers' AI conversations. Not on theoretical models, not on test environments, but on the reality of how AI models talk about them every day. This is the foundation on which it makes sense to build an AI visibility strategy. And it's the foundation on which we built Geosnap. ## Frequently asked questions ### Why isn't querying ChatGPT via API enough to measure a brand's visibility? Because the API queries the model in a controlled technical environment, without the session context and interaction mode of a real user, which influence the answer. Geosnap has seen brands that looked well positioned in API analyses but never appeared in the answers a real user received, and vice versa. ### How does Geosnap's real user simulation work? Geosnap opens the browser, accesses the AI platform (ChatGPT, Gemini or Perplexity), builds the session context as a user would, enters the question the way a person would write it and analyzes the complete answer. Prompts are generated dynamically based on the client's brand, industry and market. ### How often does Geosnap repeat its AI visibility analysis? Every month, on the same set of prompts, so the data is comparable over time and progress can be measured meaningfully.
Structured data
Schema.org data (JSON-LD) embedded in the page code. It tells the AI explicitly what the page is about and who published it.
BreadcrumbList describes the path through the site
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Home",
"item": "https://geosnap.ai/en/"
},
{
"@type": "ListItem",
"position": 2,
"name": "Blog",
"item": "https://geosnap.ai/en/blog/"
},
{
"@type": "ListItem",
"position": 3,
"name": "Product",
"item": "https://geosnap.ai/en/blog/#prodotto"
},
{
"@type": "ListItem",
"position": 4,
"name": "We Dropped the APIs and Now Simulate the Real User",
"item": "https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale"
}
]
}
BlogPosting describes the article: author, dates and sources
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"@id": "https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale#article",
"headline": "We Dropped the APIs and Now Simulate the Real User",
"description": "Why Geosnap doesn't measure AI visibility through APIs: it simulates real users on ChatGPT, Gemini and Perplexity to capture the answers customers see.",
"inLanguage": "en",
"url": "https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale",
"mainEntityOfPage": "https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale",
"image": "https://geosnap.ai/img/tOioVlHoERulNIypDeKYg3D70.webp",
"articleSection": "Product",
"wordCount": 689,
"author": {
"@type": "Person",
"name": "Rinald Sefa",
"jobTitle": "CMO",
"url": "https://geosnap.ai/en/about",
"worksFor": {
"@id": "https://geosnap.ai/#organization"
}
},
"publisher": {
"@id": "https://geosnap.ai/#organization"
},
"isAccessibleForFree": true,
"abstract": "Geosnap doesn't query AI models via API, because API calls run in a controlled technical environment that doesn't reflect what a real user sees and can produce partial or misleading data. Instead, it simulates the full interaction on ChatGPT, Gemini and Perplexity, from the browser and session context to the question itself, and repeats the process every month on the same prompts to get data that's comparable over time.",
"translationOfWork": {
"@id": "https://geosnap.ai/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale#article"
}
}
FAQPage describes the frequently asked questions
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "https://geosnap.ai/en/blog/abbiamo-abbandonato-le-api-e-simuliamo-l-utente-reale#faq",
"mainEntity": [
{
"@type": "Question",
"name": "Why isn't querying ChatGPT via API enough to measure a brand's visibility?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Because the API queries the model in a controlled technical environment, without the session context and interaction mode of a real user, which influence the answer. Geosnap has seen brands that looked well positioned in API analyses but never appeared in the answers a real user received, and vice versa."
}
},
{
"@type": "Question",
"name": "How does Geosnap's real user simulation work?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Geosnap opens the browser, accesses the AI platform (ChatGPT, Gemini or Perplexity), builds the session context as a user would, enters the question the way a person would write it and analyzes the complete answer. Prompts are generated dynamically based on the client's brand, industry and market."
}
},
{
"@type": "Question",
"name": "How often does Geosnap repeat its AI visibility analysis?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Every month, on the same set of prompts, so the data is comparable over time and progress can be measured meaningfully."
}
}
]
}
Metadata
The information in the page header, which crawlers and engines read before the content.