Maya — Overview

Quick start (for engineers)

Want to skip the marketing? In order:

1. Pick your stack — IIS, Nginx, Apache, Cloudflare, AWS, or a custom BFF endpoint.
2. Apply the source-side filter — the allowlist + denylist your filter must produce.
3. POST a test batch — X-Maya-Mode: test validates without ingesting.
4. Add markdown rendering — optional but high-ROI; gated by feature flag.

If you have 10 minutes, run the IIS or Nginx filter in --dry-run mode and POST a sample to /v1/logs with X-Maya-Mode: test. That's the entire integration loop.

What Maya is

Maya is a platform that measures and improves how a brand is represented inside Large Language Models (LLMs) such as ChatGPT, Gemini, Claude, Perplexity, and Copilot.

Maya operates on three layers:

1. Bot log analysis — ingestion of server-side logs to identify how LLM crawlers interact with your site.
2. Prompt simulation — automated, scheduled queries to LLMs to measure your brand's mention rate, position, and competitor co-occurrence.
3. Markdown rendering optimization — recommendations and reference implementations for serving LLM bots high-density, low-overhead content.

What Maya is NOT

Maya is intentionally not a replacement for, and not a competitor to:

• Web analytics platforms (GA4, Adobe Analytics, Matomo, Dataroid). Maya does not track end users.
• CDPs / customer data platforms. Maya does not ingest, store, or process customer data.
• SEO crawlers (Screaming Frog, Sitebulb). Maya does not crawl your site; it observes how external bots crawl it.
• Marketing automation tools. Maya does not send campaigns, push notifications, or messages to end users.

Who Maya is for

• Mid-market and enterprise brands whose customers are increasingly discovering products through LLMs.
• Regulated industries (banking, insurance, healthcare, legal) where data minimization is non-negotiable.
• Marketing teams that already have GA4 / Dataroid coverage on the user side and need symmetric coverage on the bot / model side.

How Maya is different

Question	Answer
Does Maya touch user data?	No. Server logs are filtered before transmission so only verified LLM bot traffic reaches Maya.
Does Maya modify our site?	No. Markdown rendering is a self-implemented optional feature behind a feature flag controlled by you.
Does Maya use our data to train models?	No. Your data is used only to populate your dashboards.
Is data retained by upstream LLM providers?	No. Maya uses enterprise-grade endpoints with zero-retention guarantees.
Where does Maya host data?	Türkiye-only. Maya is a Turkish company; all tenant data is hosted and processed on servers in Türkiye. Encrypted at rest (AES-256) and in transit (TLS 1.3).

Core concepts at a glance

• LLM bot: an automated agent operated by an LLM provider (e.g., GPTBot, ClaudeBot).
• Prompt set: a curated list of queries Maya runs against LLMs on your behalf, weekly or daily.
• Mention: an instance of your brand appearing in an LLM's response.
• Citation: a URL the LLM links to as a source.
• Mention rate: percentage of prompts in a set where your brand is mentioned.
• Markdown rendering: serving a markdown-formatted response to verified LLM bots, when (and only when) they request it.

For a full glossary, see getting-started/concepts.md.

Engagement model

A typical Maya engagement starts with the following milestones. Times are indicative.

Week	Milestone
1	Account provisioned. Prompt set loaded. Search Console + GA export connected.
2	Joint technical meeting with the brand's IT / engineering team. Log integration path agreed.
3–4	First sanitized log batch. First insights report.
5–8	Markdown rendering implementation (optional, brand-implemented behind a feature flag).
9+	Ongoing measurement, monthly insights, expansion to additional brands or markets.

Next steps

• Engineering: integrations/log-export/overview.md
• Compliance: security/overview.md
• Product: integrations/markdown-rendering/overview.md