Prompts

Prompts tracking overview

How AIclicks tracks the questions your buyers ask AI engines

In the app, view and manage your prompts in Prompts, explore new ideas in Discover prompts, and inspect variations in Query fan outs.

What a "prompt" means

In traditional SEO, you track keywords. In AI search, you track prompts.

A prompt is a question or query that a real user types into an AI engine like ChatGPT, Perplexity, or Gemini.

Examples:

What are the best hotels in Lisbon for business travelers?

These are the actual inputs buyers use to discover products, compare options, and make decisions.

AIclicks runs those prompts across 10+ AI platforms and captures what each engine says. Your brand either appears in the answer or it doesn't.

Prompts vs. keywords

Keywords and prompts look similar but work differently.

A keyword is a short search term. A prompt is a full question with intent and context built in.

Keywords	Prompts
best travel booking site	What's the best site to book last-minute flights in Europe?
hotel management software	Which hotel management software works best for boutique properties?
travel insurance	best travel insurance I should get for a 3-month trip to Southeast Asia

Prompts are longer, conversational, and include specific constraints. This matters because AI engines don't match keywords. They analyze the entire question to understand what the person actually needs, then synthesize an answer from trusted sources.

When you track prompts, you're tracking what your buyers are literally typing into AI engines, not abstract search terms.

How AI engines process prompts

AI engines like ChatGPT analyze three things when a user asks a question:

Intent. What the user is actually asking for. "What's the best hotel booking platform?" and "Which site should I use to book hotels?" carry the same intent. AI engines recognize this and produce similar responses to both.
Context. The constraints and situation details in the question. "For budget backpackers" produces a different response than "for business travel with expense reporting." The context shapes which sources the engine pulls from and which brands it names.
Response generation. The engine synthesizes sources it considers reliable and produces one answer. Not a list of links. One answer that names a few brands.

This has a practical implication for tracking: you don't need to track every possible variation of the same question. AI engines recognize similar intent across different phrasings. What matters is that you capture the right intent and context, not that you perfect every word.

How AIclicks tracks prompts: API vs. UI scraping

There are two ways to pull data from AI platforms: via their developer API or by querying them the same way a real user would. AIclicks uses the second approach.

API-based tracking	UI-based tracking
It sends a request directly to the model's backend. It's fast and structured, but it bypasses the full user experience. APIs often use older model versions, skip real-time web retrieval, and miss the Retrieval-Augmented Generation (RAG) layer that shapes what sources get cited in live answers. The data you get reflects what the model knows in isolation, not what it actually tells users.	It queries each AI platform through its live interface, exactly how your buyers use it. This captures the complete response pipeline: web retrieval, source selection, real-time synthesis, and final output. What AIclicks records is what a real user would see.

This distinction matters for three reasons:

Accuracy. AI engines like Perplexity and ChatGPT with browsing enabled actively retrieve web sources before generating answers. API calls often skip this step. UI scraping captures the full retrieval layer, so the brands and citations you see in AIclicks match what users actually get.
Citation data. The sources AI engines cite come from the live web retrieval layer, not the base model. UI scraping gives you real citation data. API tracking often gives you none.
Recency. Live interfaces reflect the latest model updates and retrieval behavior. API versions frequently lag behind. With UI-based tracking, you're always seeing the current user experience, not a frozen snapshot.

The result is prompt data that's closer to ground truth. What you track in AIclicks is what your buyers see.

What you can do with prompt data

Prompt tracking gives you three types of signal:

Visibility signal. Which prompts does your brand appear in? Where are you missing? What's your share of voice against competitors on specific questions? Learn how this rolls up into the Visibility metric and Share of Voice.

Content signal. What topics, formats, and sources are AI engines citing when they answer prompts in your category? This tells you what to write and where to get mentioned. Read more in Citations (Frequency) and Sources Overview.

Competitive signal. Which prompts are competitors winning that you aren't? Which AI platforms favor them over you? Where is there a gap you can close?

What comes next

There are two concepts that sit directly on top of prompt tracking:

Create prompts

How to choose prompts, how many to track, how to read the data, and how to act on it.

Query fan out

Learn why the same intent generates dozens of different prompt variations and how to leverage this information.

Was this page helpful?

Last updated today

Built with Documentation.AI