Top Competitors & sentiment

The Top Competitors card on your dashboard shows the brands AI engines are mentioning instead of (or alongside) yours when they answer category-relevant questions.

How competitors get detected

Every Monday’s monitoring run produces ~50 LLM responses (10 queries × 5 LLMs). Claude Haiku reads each response and extracts every brand mentioned, with metadata:

• Sentiment (positive / neutral / negative) toward that brand
• Position (primary recommendation / mentioned in body / passing reference)
• Sample context (a quote showing the brand in context)
• Co-mentioned brands (the other brands mentioned in the same response)

Brands that show up across multiple responses become “Top Competitors.” A clustering pass collapses spelling variants (e.g., “Ella + Mila” / “Ella+Mila” / “ellamila” → one canonical entry).

You can also pre-supply Known Competitors in onboarding. Those names get merged into the analysis so brand-family relationships and odd spellings get detected correctly.

What each metric means

For each Top Competitor, the card shows:

Mentions

Total count across the 50 responses. Higher = the LLMs are more likely to surface them when answering category questions.

Positive sentiment %

Share of mentions where the LLM described the brand favorably. Important caveat: this is sentiment toward the COMPETITOR, not toward you. A 95% positive rating on a competitor means the LLMs are describing them in glowing terms — which is the gap your content needs to close.

Positioned as primary recommendation %

Share of mentions where the LLM positioned that competitor as its top suggested answer, not just a passing reference.

This is the metric that separates “name-dropped” brands from “actually-recommended” brands. A brand with 50 mentions and 10% primary rec is mentioned a lot but rarely the LLM’s go-to. A brand with 30 mentions and 80% primary rec is much more dangerous — they’re displacing you at the moment of decision.

Sample context

A real quote from one of the LLM responses, showing how the brand was described.

Often co-mentioned with

The 3 brands most frequently mentioned alongside this one.

The “Sort by Recommended First” toggle

By default the card sorts competitors by raw mention count. Toggling to “Recommended First” sorts by mentions × primary_rec_pct — the brands actually winning recommendations, weighted by mention volume.

This is usually the more actionable view. Pure mention count rewards brands name-dropped 50× across all queries. Position-weighted reveals which brands are the LLMs’ “go-to recommendation” for your category — the ones you actually need to displace.

Competitive Sets

Below the Top Competitors grid, the “You’re being compared against these competitive sets” section shows clusters of brands that frequently appear together in LLM responses. Each set is a chip-row of brand names.

Why this matters: the set an LLM mentions you alongside reveals the category framing. If you’re in Set 1 (premium nail polish brands), you’re being positioned alongside the premium tier. If you’re absent from every competitive set, that’s the diagnosis — you’re not part of the conversation at all.

The most common pattern: customers in their first month appear in zero competitive sets, then by month 3 they appear in 1–2 sets as their schema and content depth get consistently picked up by the LLMs.