About AICite

Who's behind it

AICite was created by Hector Londoño (@Heloqui on GitHub) in April 2026. Solo founder, full-stack engineer. Before AICite, Hector built multiple SEO sites and content tools — he was auditing them manually when he noticed that ChatGPT and Claude weren't citing them despite solid Google rankings. That observation became the product.

No team. No investors. AICite is an experiment to validate whether a single person, using modern AI as leverage, can build a B2B SaaS from scratch. Every decision is documented publicly: specs, implementation plans, audit snapshots, cost-per-audit breakdowns — all in the public repo.

Methodology behind the audits

Each audit runs a 5-stage pipeline validated internally:

1. Page ingestion— public HTML fetch, extraction of title, headings, existing schema.org markup, and body text. No headless browser, no JS execution — AI crawlers don't run JS either, so we see exactly what they see.
2. Query generation — an LLM (gpt-4o-mini with structured outputs) generates 15 plausible queries a user would ask an AI when looking for your niche. A mix of informational, commercial, navigational, and long-tail.
3. Multi-engine querying — parallel requests to ChatGPT (gpt-4o with web search) and Claude (sonnet-4-6 with web search). Promise.allSettled for resilience against partial failures.
4. LLM-as-judge citation extraction — Claude Haiku 4.5 (cheap and consistent for structured classification) reads each response and extracts whether your site was cited, at what position, with what sentiment, and who was cited instead. Combined with deterministic URL matching against sources.
5. Fix generation — Claude Sonnet 4.6 with forced tool_use compares your content against the cited pages and proposes specific fixes — exact text to add, missing schema, recommended sections.

Total: ~$1.50–3.00 USD in compute per audit. Time: 30–60 seconds. The tool audits itself — the first audit of aicite.app, before implementing any fixes, scored 0/100. One hour after implementing 3 fixes (schema, content, glossary), a re-audit rose to 27/100. That recursive validation is public and reproducible: see the result.

Transparent commitments

• Build in public — code on GitHub, product decisions on Twitter, metrics (signups, MRR, churn, AI costs) shared publicly every week.
• No ad tracking — only essential cookies (session, rate-limiting). We will never sell your data. We will never add Meta or Google Ads pixels.
• Real, permanent free plan— 1 audit per IP per day, with shareable public results. Not a trial, not a hook — it's the product's marketing tool, and it will stay that way.
• No-questions refunds— if you pay and it doesn't work for your use case, full refund within the first 14 days. No form, just email.
• Exportable data — everything you produce in AICite (audits, fixes, configurations) can be exported as JSON whenever you want, no permission requests needed.

Tech stack (for the curious)

Next.js 16 (App Router) on Vercel · Turso (libSQL) as the database · Trigger.dev v4 for pipeline background jobs · OpenAI gpt-4o + gpt-4o-mini · Anthropic Claude sonnet-4-6 + haiku-4-5 · Auth.js v5 (magic link + Google OAuth) · Stripe billing · Resend for emails · Tailwind CSS 4 · class-variance-authority · zod for validation · Vitest for tests · Sentry for errors. All TypeScript strict.

The repo is public for auditing: github.com/Heloqui/aicite.

Contact

For support, bugs, feedback, or legal requests (GDPR, CCPA), email hello@aicite.app. For public conversation or build-in-public updates, follow the project on GitHub.