Research, methodology, and longitudinal observations from inside the observatory. We share what we see, why we think it matters, and the tools we use to see it more clearly.
When someone asks an AI for “the best espresso machine for a beginner”, the answer doesn’t name a brand — it names a model. BotScope now tracks which specific products AI answers recommend, with per-model sentiment, plus ChatGPT’s shopping carousel: the prices it shows and the merchants it sends buyers to.
We ran 384 ChatGPT ad observations across four categories OpenAI bans outright, to test whether prohibited ads slip through. Mostly they do not. What runs instead is a playbook worth stealing — including one advertiser clearing a gambling ban by not technically being gambling.
The models we track fail sometimes. When they do, the dangerous outcome is not a missing number — it is a plausible one. Here is how BotScope reports a partial scan, and how it now repairs one without being asked.
The same question gets a different answer in London than it does in Chicago — different brands recommended, a different field of competitors, sometimes your brand missing entirely. BotScope projects now track a market, and every scan is observed from it.
OpenAI put ads inside ChatGPT answers, and there is no public ad library to look them up in. BotScope now tracks every sponsored placement that appears on your prompts — who is advertising, what they say, and whether they are buying answers you do not appear in.
Averages hide the story. Tag your tracked prompts — by topic, audience, or a campaign label you invent — and every chart, report, and API call can be filtered to just the work you want to measure.
BotScope now has a read-only REST API and a hosted MCP server. Pull your visibility data into your own stack — or just ask Claude how your brand is doing and let it query BotScope directly.
Your visibility charts show what moved. Annotations capture why — content changes, prompt edits, provider switches, model updates — pinned to the exact date, and ready to show clients.
An AI visibility tool tracks whether models like ChatGPT and Claude mention, recommend, and cite your brand. Here are the nine best in 2026 — including ours, clearly labelled.
SEO earns you a position in ranked results. GEO earns you a place inside AI-generated answers. Here is what actually changes — surfaces, tactics, KPIs — and what stays exactly the same.
A new generation of customers asks Claude, ChatGPT, and Perplexity instead of Google. We needed an instrument to watch what the models were saying.
APIs have been the backbone of software integration for decades. Now the Model Context Protocol (MCP) is emerging as a new standard built specifically for AI agents. How do they compare, and when should you use each?
The Model Context Protocol (MCP) is quickly becoming one of the most talked-about standards in AI. But what exactly is it, how does it work, and why should you care? This guide breaks it all down in plain language.
With the rapid evolution of AI technologies, methods such as Retrieval-Augmented Generation (RAG) and Model Context Protocol (MCP) are becoming central in improving AI-generated content. While both methods aim to enhance AIs ability to provide accurate and contextually relevant outputs, they operate in distinctly different ways. Lets look into the key differences and their implications.
The rise of AI agents—autonomous systems that can interact with digital environments on behalf of users—is reshaping the internet as we know it. As these agents become more prevalent in how people search, shop, and interact online, **website design must evolve** to accommodate this new kind of "user."
In the fast-moving world of AI, new technologies and protocols are emerging that promise to reshape how machines interact with the web. One of the most exciting developments is the **Model Context Protocol (MCP)** — a standard designed to improve how AI models understand, navigate, and retrieve content from websites.