AI visibility

A visibility project tracks one brand across the AI models. Create one under AI Visibility → New project, enter the brand name and website, and add the competitors you want to benchmark against. CrunchJunky uses the brand and competitor names to detect mentions in AI responses, so spell them exactly as they appear in the market — including common variations if the brand is known by more than one name. Each project is independent, which makes it easy to run visibility for several clients side by side. Once the project exists, you add prompts and choose which models to scan, then CrunchJunky takes over the daily monitoring.

Prompts are the questions CrunchJunky asks the AI models on your behalf. Add them under the project's Prompts tab and tag each one by intent: Discovery prompts are unbranded category questions ("best marketing reporting tools for agencies"), Brand prompts name your client directly ("is Lumière any good?"), and Competitor prompts are comparisons or alternatives ("Lumière vs Velora"). A good starting set mixes all three categories — perhaps fifteen to thirty prompts that mirror how real buyers search. CrunchJunky runs every prompt against every enabled model on each scan and records whether your brand was mentioned, where it ranked in the answer, how it was described, and which sources the model cited.

CrunchJunky can track eight AI surfaces: ChatGPT, Gemini, Perplexity, Claude, Google AI Overviews, Google AI Mode, Grok, and DeepSeek. Enable the ones that matter for your client's audience under the project's Models settings — you don't have to run all eight. Each enabled model is scanned automatically every day. Some models require specific setup. Google AI Overviews, for example, is tracked through a search-results provider rather than a chat API. The Models settings page tells you exactly what each one needs and shows the current status of every connection at a glance.

Four headline metrics tell you how you're doing. Visibility % is how often your brand appears across all tracked prompts. Share of Voice is your slice of total brand mentions versus your tracked competitors — the clearest signal of who owns the conversation. Sentiment scores how positively each mention describes you, and Average Position records where you land when you do appear. Drill into any metric by model, by prompt category, or over time. The Responses view lets you filter to All, Mentioned, Not mentioned, or Mention gap (prompts where a competitor appears and you don't) — the fastest way to find concrete opportunities. The Sources view ranks the domains AI cites most, and the Insights view turns all of this into prioritised, plain-language actions.

When an AI model answers a prompt it often cites the pages it drew on, and CrunchJunky records every one. The Sources view ranks the domains cited most across all your prompts and models — the sites that are actually shaping what AI says about your category. The URLs view goes a level deeper, listing the individual pages cited and how often each appears, so you can see exactly which articles, listicles or product pages the models lean on. Gap analysis turns this into a to-do list. It surfaces the domains where AI cites a competitor but never you — your clearest content gaps. Each gap is a concrete opportunity: earn a mention on that source (a review site, a directory, an industry roundup) and you start appearing where buyers are already being pointed. Because all three views are built from real scan citations, they update every time a scan runs.

Before an AI can mention a brand, its crawler has to be allowed to read the site. The Crawlability tool checks a domain's robots.txt against every major AI crawler — GPTBot and OAI-SearchBot (ChatGPT), ClaudeBot and Claude-SearchBot, PerplexityBot, Google-Extended and Googlebot, and more — and tells you whether each is Blocked, Partial or Allowed. Expand any crawler to see exactly why (which user-agent rule matched) and how to change it. The tool distinguishes training crawlers (blocking them keeps your content out of model datasets, often intentional) from search and answer crawlers (blocking these directly hurts your AI visibility), so you can make the right call per bot. The domain pre-fills from the client's saved website. Crawl insights is the companion view: once you forward your server logs, it shows how often each AI crawler actually visits the site — turning "we allow GPTBot" into "GPTBot fetched 40 pages last week". Together, Crawlability tells you who *can* read the site and Crawl insights tells you who *is*.

Larger projects benefit from a little organisation. The Brands view manages the exact names CrunchJunky matches as mentions — your client plus its competitors — including spelling variations and aliases, so a mention is never missed or miscounted. Tags let you label prompts (by theme, funnel stage, or client priority) and then filter every metric by tag, which is the fastest way to answer questions like "how visible are we on bottom-of-funnel comparison prompts?". Project Settings is where you control the brand name and website, the competitor list, which models run, runs-per-prompt, and the scan schedule (off, daily, or weekly). Changes here apply from the next scan onward.

Already tracking AI visibility in Peec AI? You can import a Peec project straight into CrunchJunky instead of running native scans for that client. Under Data → Integrations, open Connect on the Peec AI card, choose the client, and paste a project-scoped Peec API key (created in Peec under Account → API Keys). CrunchJunky verifies the key, then imports the project's visibility, share of voice, sentiment, average position, prompts and cited sources. Once imported, the data populates the exact same AI Visibility screens and report widgets as native scans — Overview, Prompts, Sources, Gap analysis and the report builder all work unchanged. The project is marked as Peec-sourced so the origin of the numbers is always clear, and re-syncing pulls the latest figures. Note that Peec's REST API is currently part of their Enterprise/beta tier, so the key needs API access enabled on their side.

You can run scans two ways. Bring-your-own-keys lets you connect your own API keys for each model; CrunchJunky runs the scans through them and you pay only the providers' usage rates, with no markup. Keys are encrypted at rest and used solely for your scans. This is the most cost-effective option if you already have provider accounts. Managed AI is the no-setup alternative: for a flat €19 per model per month, CrunchJunky runs the scans on our own infrastructure — no keys to manage, no usage to forecast. You can mix the two approaches, turn individual models on or off at any time, and switch between BYO and Managed per model. See the pricing page for current Managed AI details.