What we believe.

AI assistants — ChatGPT, Claude, Gemini, Perplexity — replaced the ten blue links with two to seven cited sources per answer. For brands, that compresses two decades of search-funnel optimization into one question: are you in the citation set, or are you not?

Most platforms in this category answer that question with a dashboard. They tell you you're missing. They charge $500 to $2,000 a month for the privilege of watching your invisibility in real time. None of them write the code that puts you back on the map. Findrix does.

Detection without a written fix is a slide deck. The defensible thing in this category is the fix-writing layer — schema, content rewrites, citation patches, authored per stack and ready to paste — across the four stacks the long tail of mid-market actually uses (WordPress, Shopify, Next.js, Webflow).

Getting 40 CMS dialects right is unglamorous and slow. That moat compounds. Every fix we write makes the next one cheaper. Every stack we shake down on a real client site adds reliability we sell.

We report Wilson confidence intervals on every percentage. We run BCa Bootstrap on share-of-voice. We apply Benjamini-Hochberg FDR correction on prompt batteries. Every reported number carries a method ID a CFO can trace.

Vectory publishes the formulas. We ship the formulas, the calculators, and the open-source library that reproduces every number. Read the methodology page for the math.

Every audit improves the next one. We learn which schema patches actually move citations on which stack, which apply-to-lift latency to expect per LLM, which prompt families regress to the mean fastest. That meta-data — how the AI citation surface actually responds to interventions — is the moat we're building.

We don't train models on customer data. We don't sell audit data to LLM providers. The flywheel runs on aggregated outcome statistics with strict tenant isolation. See our privacy policy.

• Prompt injection. No hidden instructions, no Unicode tricks, no jailbreaks. We write legitimate on-domain content and standards-compliant schema only.
• Content spam. No mass-generated pages, no LLM-rewritten posts at scale, no link farms. Every fix is reviewable in your CMS or your git.
• Cloak the math. If our methodology can't survive a peer review, it doesn't ship to a customer report. Open formulas, open assumptions, open limits.
• Sell to LLM providers. Customer audit data stays in customer tenants. We share aggregated, anonymized outcome stats only — and only with explicit consent.

General: team@findrix.ai

Agencies: agencies@findrix.ai

Press / partnership: press@findrix.ai

Security: security@findrix.ai (see security.txt)

• Wordmark + lockups (SVG, PNG): press-kit.zip (Phase 6)
• One-paragraph boilerplate (under brand story above)
• Founder bios + headshots: available on request via press@findrix.ai