Tracking whether your brand gets mentioned in AI responses is only half the story. The other half — arguably the more actionable half — is which URLs the AIs are actually sending users to. Today we're shipping the biggest update to AI Visibility since launch.

🔗 New: URLs & Rankings Tab

For every tracked keyword, you can now see exactly which URLs each AI model cites in its response — automatically classified:

• Your URLs — links pointing back to your own domain. These are gold; the AI equivalent of an organic #1.

• Competitor URLs — pages on the competitor domains you've configured. This is your visibility gap, made measurable.

• Other sources — third-party domains the models referenced (review sites, marketplaces, comparison articles).

Each domain expands to show every distinct URL cited, with model badges showing exactly which AI linked to which page. Hover to see the full path. Click to open it.

📊 Why This Matters

When a user asks ChatGPT "what's the best CRM for small business," the model doesn't just give a list — it often links to specific product pages, comparison articles, or review sites. Until now, the only way to find those URLs was to read through 8 AI responses by hand.

Now you get a domain-grouped, model-attributed view at a glance — plus a Brand URL Coverage stat that shows what fraction of the 8 models actually linked to your domain at least once.

⚡ Inline Keyword Analysis (No More Modal)

Click any keyword in your project — the row expands in place to reveal:

• A glanceable strip of 8 model logos with citation-status dots

• AI Responses — the raw text each model returned

• URLs & Rankings — the new domain/URL view above

• Insights — performance summary + improvement ideas

Multiple keywords can be open at once. No popup, no scroll-lock. Click again to collapse.

🔄 On-Demand Re-tracking

You no longer have to wait until Sunday's automatic sync to refresh a keyword. Each keyword has a 2× per day re-track quota — click the refresh button on any row, or use Re-track All at the project level to refresh everything that has quota remaining.

The 8-model query typically completes in 10–30 seconds; the page polls automatically so fresh results appear without reloading.

The weekly Sunday sync runs exactly as before — the daily quota only applies to manual button clicks, not to the automatic schedule.

🤖 Refreshed Model Lineup

We've updated the tracked models to the latest catalog:

• Claude Sonnet 4.6 — Anthropic

• GPT-5.5 — OpenAI (new)

• Gemini 3.1 Pro — Google

• Gemini 3.5 Flash — Google (new)

• Gemini 2.5 Flash — Google

• DeepSeek V4 Pro — DeepSeek (new)

• Llama 4 Maverick — Meta

• Mistral Large 3 — Mistral AI

Historical tracking data remains fully viewable — model labels gracefully fall back to legacy IDs where needed.

🧭 A Note on Methodology

We deliberately do not enable web-search-augmented model calls on these queries. The whole point of AI Visibility is to measure what the models know from their training data — that's what a real ChatGPT or Claude user gets when they ask a question without explicitly turning on search.

Enabling live web search would turn this into a Google-ranking proxy. The citations you see come from the model's organic memory, prompted to prefer specific product or category pages when known, with strict instructions not to invent URLs.

💡 What You Can Do With This

• "Are any of the 8 AIs actually linking to our product page, or just our homepage?"

• "Which competitors are getting linked when users ask AIs about our category?"

• "Did our recent llms.txt + structured-data work move the needle? Re-track now and compare."

• "Show me a third-party domain that every model cites — we should reach out to them."

• "Which keywords are at 0/8 brand URL coverage? Those are the AI-SEO priorities."

AI search is rewriting the rules of discovery — and URL citations are the first concrete, measurable signal of what the AIs are actually showing users today. This is the most actionable view of AI Visibility we've shipped yet.

— The Zutrix Team