JD
All projects

AI Visibility Platform

AI View Sync

See how AI models talk about your brand — then fix it, automatically.

Founder — product, engineering, and go-to-marketLive product

01

Problem

Buyers increasingly ask AI assistants — not Google — what product to use. Most companies have zero visibility into whether ChatGPT, Claude, Gemini, or Perplexity recommend them, mention them accurately, or recommend a competitor instead.

SEO tooling measures rankings on a results page that fewer buyers see every quarter. There was no equivalent of Search Console for the AI layer: no audit, no share-of-voice metric, no way to know if your content is even legible to a language model.

Even teams that understood the problem had no path from insight to action. Knowing an LLM misdescribes your pricing is useless if fixing it means a quarter-long content project.

02

Solution

AI View Sync runs structured audits against multiple frontier models, asking the questions real buyers ask and scoring how each model represents the brand — accuracy, sentiment, recommendation rate, and share of voice against named competitors.

It then closes the loop: the platform generates llms.txt files, FAQ schema, and content recommendations mapped to the gaps it found, and syncs approved changes directly to WordPress. The audit isn't a report — it's a queue of fixes with an apply button.

A growth roadmap ties it together, sequencing the highest-leverage fixes first so a marketing team can improve AI visibility the way they'd run a CRO program: measure, change, re-measure.

Capabilities

What it does

AI Recommendation Audit

Structured prompt batteries simulate real buyer questions across models and score whether — and how — the brand gets recommended.

LLM Visibility Analysis

Measures how accurately each model describes products, pricing, and positioning, and flags hallucinated or stale claims.

Competitor Analysis & Share of Voice

Tracks recommendation rate against named competitors over time — the AI-era equivalent of rank tracking.

AI Traffic Reporting

Segments referral and assistant-driven traffic in GA4 so teams can see what AI visibility is actually worth.

llms.txt & FAQ Schema Generation

Generates machine-readable brand context and structured data so models cite the source of truth instead of guessing.

WordPress Sync & Automated Implementation

Approved recommendations publish directly to the site — schema, content updates, and llms.txt — without a developer in the loop.

Content Recommendations

Maps every visibility gap to a concrete content change, prioritized by expected impact on recommendation rate.

Growth Roadmap

Sequences fixes into a measurable program: baseline audit, implementation sprints, and scheduled re-audits.

Architecture

How it's put together

The platform is a Next.js app over a job-queue backend. Audits fan out across model APIs in parallel, results are normalized into a scoring schema, and an implementation engine turns findings into deployable artifacts.

Client

Next.js Dashboard

Audits, SoV, roadmap

Reports

Scheduled PDF / email

API & Orchestration

REST API

Auth, orgs, projects

Audit Queue

Fan-out job runner

Scheduler

Re-audit cadence

AI Layer

Prompt Battery Engine

Buyer-intent question sets

Claude · GPT · Gemini · Perplexity

Parallel model calls

Scoring & Extraction

Accuracy, sentiment, SoV

Implementation

llms.txt Generator

Schema Builder

FAQ / Organization JSON-LD

WordPress Sync

REST API publisher

Data

Postgres

Audits, scores, history

GA4 / BigQuery

AI traffic reporting

Challenges

What was hard, and what I did about it

LLM answers aren't deterministic

The same question yields different answers run to run. Single-shot audits were noise. The fix: repeated sampling per question, answer clustering, and confidence intervals on every score — treating model output as a distribution, not a fact.

Scoring 'how a model talks about a brand'

Recommendation rate is easy; accuracy and sentiment are not. I built an extraction layer where a second model grades answers against a structured brand fact sheet, with spot-check tooling to keep the grader honest.

Automated publishing people can trust

Nobody lets a tool write to their production site on day one. Every change ships through a review queue with diffs and one-click rollback, which turned 'automated implementation' from scary to adopted.

Tech stack

Built with

Next.jsTypeScriptTailwind CSSNode.jsPostgresClaude APIOpenAI APIGemini APIPerplexity APIGA4BigQueryWordPress REST API

Outcomes

What changed

  • 4 modelsaudited in parallel per brand, with per-model share-of-voice tracking
  • Minutesfrom audit finding to published fix via WordPress sync — previously a content-team quarter
  • Repeatableaudit → implement → re-audit loop that turns AI visibility into a measurable channel

Future roadmap

Where it goes next

Agentic audits: models that browse a site the way an assistant's retrieval layer does, not just answer questions

Shopify and headless CMS sync targets alongside WordPress

Alerting when a competitor's recommendation rate moves against you

Public AI visibility benchmarks by industry

Next project

Heynet