Skip to main content
Find answers to common questions about Lightsage, how AI Engine Optimization works, and what you can do with the platform. If you don’t see what you’re looking for, reach out through the community or book a demo.

​
General

GEO (Generative Engine Optimization) is the practice of improving how often and how positively AI answer engines like ChatGPT and Perplexity mention your brand in their responses. AEO (AI Engine Optimization) goes further — it covers action engines, specifically AI coding agents like Claude Code, Codex, and Cursor that write and execute code on behalf of developers.For API and developer tool companies, AEO matters more than GEO. Getting mentioned in a ChatGPT response is brand awareness. Getting recommended and successfully used by Claude Code is developer adoption.
Consumer GEO tools like Profound and Otterly are designed to track whether AI answer engines mention your brand in conversational responses. That’s useful for consumer products — if ChatGPT mentions your app when someone asks “what’s the best productivity tool?”, that’s a win.For APIs and developer tools, brand mentions don’t translate to adoption. What matters is whether a coding agent like Claude Code recommends your SDK when a developer asks it to build something, and whether the generated code actually works. Consumer GEO tools don’t track coding agents at all, and they have no way to measure tool call success rates, error recovery, or task completion. Lightsage is built specifically for that layer — the action engine layer where developer adoption actually happens.
Lightsage supports 17 AI platforms across three categories:Coding agents (10): Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, OpenClaw, OpenCode, Hermes, Pi, KiloAnswer engines (7): ChatGPT, Google AI Overviews, Google AI Mode, Gemini Search, Perplexity, Claude, Microsoft CopilotModels for API Performance (4): DeepSeek, Kimi, Z.ai, GrokThe Free plan gives you access to 1 platform. The Pro plan unlocks all 17.
Devtool Arena is Lightsage’s free, public leaderboard that ranks APIs and developer tools by how well AI coding agents can discover, call, and use them. Rankings are based on real benchmarks across discoverability, tool call success rates, and error rates — scored across multiple LLMs.Any team can check their Devtool Arena ranking on the Free plan. It’s a quick way to see where you stand against competitors before investing in deeper tracking.
Visibility Analytics tracks how often and how favorably AI platforms mention or recommend your brand in response to prompts. It measures share of voice, average position, and sentiment across answer engines and coding agents.API Performance goes deeper: it measures how well AI coding agents can actually interact with your API endpoints. This includes whether the agent calls the right endpoint, whether the generated code executes without errors, how the agent recovers from failures, and how your endpoints compare against competitors on each of those dimensions.Visibility tells you if AI knows about you. API Performance tells you if AI can actually use you.

​
Getting started

Sign up at app.lightsage.com/welcome. No credit card is required. You’ll start on the Free plan with 250 starter credits and immediate access to one AI platform.From there you can:
  • Run prompts to see how AI talks about your API
  • Check your Devtool Arena ranking
  • Review your basic visibility score
  • Upgrade to Pro when you’re ready for full platform coverage
There are four common reasons coding agents overlook an API:
  1. Blocked AI crawlers in robots.txt — If your robots.txt blocks GPTBot, ClaudeBot, or similar crawlers, AI systems can’t index your documentation. Agents that do web searches during execution won’t find you.
  2. No llms.txt file — Coding agents and AI systems benefit from machine-readable context about your API. Adding an llms.txt file at the root of your domain gives agents structured information about your endpoints, SDKs, and use cases.
  3. No comparison content — Agents often encounter prompts where developers are choosing between two options. If you don’t have content comparing your API to common alternatives, agents default to whichever competitor has that content.
  4. Unclear error messages — When agent-generated code fails, the agent needs clear, actionable error messages to recover. Generic or vague errors cause agents to abandon your API mid-task or switch to a competitor.
Tracking AI recommendations for your API involves two distinct layers:Answer engine tracking covers platforms like ChatGPT, Perplexity, and Google AI Overviews. These tell you how often AI mentions your brand when developers are researching options.Coding agent tracking covers platforms like Claude Code, Cursor, and GitHub Copilot. These tell you whether AI recommends and successfully uses your API when developers are actively building.Lightsage handles both. Start by signing up for the free tier to see your current position in Devtool Arena, then set up prompt tracking for your most competitive use cases on the Pro plan.

​
Technical

Coding agents use a four-layer decision stack when choosing which SDK or API to recommend:
  1. Training data — The model’s base knowledge, which reflects what was in its training corpus. APIs with strong documentation and community presence have an advantage here.
  2. Web search — Many agents perform live web searches during execution to find current documentation, changelogs, or comparison content. What they find during these searches directly influences their recommendations.
  3. Structured context — Formats like llms.txt and MCP (Model Context Protocol) servers give agents machine-readable, structured information about your API. Agents that support these formats can access your endpoints and usage patterns directly, bypassing the need to infer from unstructured documentation.
  4. Execution feedback — Agents learn during a task. If the first API call fails with a clear, actionable error, a good agent recovers and retries. If errors are opaque or the SDK is hard to use, agents may switch to a competitor mid-task.
Optimizing at each layer — documentation quality, llms.txt, MCP support, and error message clarity — improves your recommendation rate across coding agents.
Lightsage provides a lightweight middleware that detects AI coding agents — including Claude Code, OpenCode, ChatGPT, and others — as they visit your website and documentation in real time.Installation takes one line of code and has no measurable impact on site performance. The middleware works with Next.js, Vercel, Cloudflare, Express, Astro, and Netlify.
Once installed, your Agent Analytics dashboard shows which agents are visiting, which pages they’re reading, and when.
Yes. Lightsage automatically identifies your competitors and tracks their visibility alongside yours. You’ll see share of voice comparisons, average position relative to competitors, and opportunities where you’re being outranked. The Devtool Arena leaderboard also shows public rankings across your product category.
Yes. The Free plan is $0/month and requires no credit card. It includes:
  • 3 prompts per day
  • 1 AI platform
  • Agent analytics
  • Basic visibility score
  • Devtool Arena access
  • Community support
  • 250 starter credits
Sign up at app.lightsage.com/welcome.

​
Billing

A credit is the unit Lightsage uses to measure evaluation usage. Every time Lightsage runs a prompt on an AI platform, tests an API endpoint, or evaluates a coding agent interaction, it consumes credits.
  • AI Search / Answer Engine platforms: 1 credit per run
  • Coding Agents: 5 credits per run
  • Models (API Performance): 1 credit per run
Coding agents cost more because they execute real code, make tool calls, and handle errors — a more compute-intensive process than querying an answer engine for a text response.
The Pro plan costs $100/month and includes 1,500 credits per billing cycle. Credits reset at the start of each billing period and do not roll over.With 1,500 credits, a typical usage pattern might look like:
  • 20 prompts × 2 AI search platforms × 30 runs = 1,200 credits
  • 5 API endpoints × 1 coding agent × 30 runs × 5 credits = 750 credits
If you need additional capacity before your billing cycle resets, you can purchase credit top-ups directly from your account settings.
Yes. You can upgrade from Free to Pro at any time from your account settings. Your new credits become available immediately upon upgrading.To move to Enterprise, or to downgrade from Pro to Free, contact the team through your account or book a call.
Volume discounts are available on Enterprise plans. If your organization needs higher credit volumes, multiple seats, or custom contract terms, contact sales through cal.com/lightsage/demo to discuss options.