Measured every L-template, every store, every engine.

The OS ran 1,104 local-visibility queries across OpenAI, Perplexity, Copilot, Gemini, and Serper Places, covering the full L-A → L-F template ladder for 17 stores in Bangalore and Hyderabad. The branded queries (L-E, L-F) all returned 90–100% recognition. The discovery queries (L-A, "stores like X near me") collapsed to single digits.

Simulated the ground truth: Serper Places became the source of truth.

Every store-fact AI quoted was diffed against Serper Places ground truth. The OS surfaced 86 wrong store hours, 101 wrong addresses, 14 wrong phone numbers, and 29 wrong ratings, including a Hyderabad location where Gemini answered "open until 10:30 PM" for a store that closes at 9:00.

• 4,943 LLM executions × 5 providers, all dedup-ed against ground truth
• 789-domain competitor graph synthesised in one pass
• 1,366 brand-alias variants resolved to 1,190 unique competitor brands

Shipped the fix queue: every L-template gap turned into a P1.

The Visibility Scout opened 318 LOCAL_CITATION action items (Justdial, magicpin, Shopsy, YouTube, Instagram). The Competitive Analyst opened 130 COMPETITOR_GAP items against Fashion Factory, Style Union, and Metro Shoes. The SEO Specialist queued 91 schema, GBP-review, and area-page tickets, categories: gbp_review_count_gap, local_schema_missing, area_page_missing, trust_signal_missing.

• 318 LOCAL_CITATION items: the third-party graph the brand didn’t own
• 130 COMPETITOR_GAP items: area-by-area parity work
• 91 SEO Specialist tickets: schema + GBP + area pages

Compounded into the OS: the L-template scoreboard now runs on every hybrid merchant.

The L-A → L-F coverage matrix and the Serper-Places ground-truth pass first ran for Cult. Both are now defaults for every hybrid merchant the OS onboards. The "AI hallucinates store hours" pattern alone has been observed in three of the three hybrid brands audited since, it now flags automatically in every new L-A scan.