This is a real AEO analysis of PostHog. We picked them because they won AI search in a crowded market and are now gaining ground in a second one. Every number keeps its raw AI answers.
Sample AEO report by Literally.dev
Sample AEO report: how AI engines see PostHog
How often ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews mention and cite PostHog. Plus three coding agents: Claude Code, Codex, and Cursor. This is the same report we build when you request an analysis.
An independent analysis by Literally.dev
Claude Code
Codex
Cursor- ChatGPT
- Claude
Gemini
Perplexity
Google AI Overviews
The measurement
240
captured AI answers
- 16 buyer-style prompts
- 8 AI surfaces, coding agents included
- Raw answers preserved and inspectable
- Re-measured quarterly
Sample AEO report
How AI engines see PostHog
We picked PostHog because they won AI search in a crowded market and are now gaining ground in a second one. The two sections below show both sides of that story. We measured public AI answers with the same system we use for paid work, and every number keeps its raw model answers, so you can check each one. This is what you get when you request your own analysis.
Sample AEO report
Product analytics prompts: the core market
Buyer questions in product analytics: picking a platform, finding funnel drop-offs, replacing Google Analytics, and running analytics, replay, and experiments in one tool. Amplitude, Mixpanel, Heap, and Pendo are scored as the brands that compete for these answers.
88%
Share of voice — AI assistants
13% citation rate across the five chat surfaces
94%
Share of voice — coding agents
25% citation rate across the coding agents
78%
mixpanel.com share of voice
Across the same prompts and surfaces
6/64
Cells with zero mentions
Each one is a buyer hearing about someone else
| AI assistants | Coding agents | |||||||
|---|---|---|---|---|---|---|---|---|
| Prompt | ChatGPT | Gemini | Claude | Perplexity | AI Overview | Claude Code | Codex | Cursor |
| What are the best product analytics tools in 2026?product recommendation | ||||||||
| Which analytics platform should an early-stage startup use to understand user behavior?product recommendation | ||||||||
| What is a good open-source or self-hostable product analytics tool?product recommendation | ||||||||
| How do the leading product analytics platforms compare on features and pricing?comparison | ||||||||
| What are good alternatives to Google Analytics for a SaaS product team?comparison | ||||||||
| How do I figure out where users drop off in my onboarding funnel?problem solving | ||||||||
| How can I combine session recordings, funnels, and A/B tests without stitching together three different tools?problem solving | ||||||||
| Who are the main players in product analytics for software teams?industry expert | ||||||||
- Primary recommendation
- Mentioned secondarily
- Not mentioned
- No data
Click any dot with a run count to inspect the captured AI answers, cited URLs, and highlighted mentions behind that result.
Who buyers hear about instead
Every company the captured answers named, ranked by share of voice (answers that mention them / answers captured).
- Amplitudeamplitude.com80%96/120
- Mixpanelmixpanel.com80%96/120
- Google Analytics 4analytics.google.com44%53/120
- Heapheap.io40%48/120
- Pendopendo.io33%40/120
- FullStoryfullstory.com24%29/120
- Contentsquarecontentsquare.com20%24/120
- Plausibleplausible.io17%20/120
- Matomomatomo.org16%19/120
- Umamiumami.is13%16/120
What gets cited
Domains behind the URLs cited in these answers (530 citations captured). Click a domain to inspect the URLs behind it.
| Domain | Citations |
|---|---|
| 29 | |
| 67 | |
| 24 | |
| 15 | |
| 15 | |
| 14 | |
| 14 | |
| 12 | |
| 11 | |
| 11 |
- Citations
- 29
- Citations
- 67
- Citations
- 24
- Citations
- 15
- Citations
- 15
- Citations
- 14
- Citations
- 14
- Citations
- 12
- Citations
- 11
- Citations
- 11
Where the citations come from
Every cited URL classified: the client’s own pages, competitor pages, social and community threads, institutional references, or earned media. Each surface leans differently — that changes the playbook per platform.
| Surface | Owned | Competitor | Social & community | Institutional | Earned |
|---|---|---|---|---|---|
| ChatGPT | 4 | 38 | 6 | — | 15 |
| Gemini | 1 | 12 | 2 | — | 63 |
| Claude | 4 | 14 | 3 | — | 107 |
| Perplexity | 5 | 10 | 3 | — | 110 |
| Codex | 10 | 26 | 9 | — | 25 |
| AI Overview | 2 | 4 | 6 | — | 22 |
| Cursor | 2 | 4 | — | — | 9 |
| Claude Code | 1 | 4 | — | — | 9 |
| All surfaces | 29 | 112 | 29 | — | 360 |
How the answers talk about PostHog
Across 108 answers that mention PostHog: 106 positive, 2 neutral, 0 negative.
Phrases the answers attach to PostHog
- “engineering-led teams”×3
- “open-source preference”
- “data ownership”
- “transparency”
- “option to self-host”
- “combines product analytics with session replay, feature flags, and experimentation in one stack”
- “open-source”
- “developer-first suite”
- “bundles analytics, session replay, and feature flags”
- “can be self-hosted or run in the cloud”
- “control, extensibility, and consolidating tools”
- “engineering-led teams and data ownership”
- “startups and engineering-heavy teams, especially those wanting self-hosting”
- “all-in-one open-source stack”
- “massive growth in 2026”
- “top choice for teams that want to own their data and avoid 'tool sprawl'”
- “analytics, feature flags, session recording, and heatmaps in a single platform”
- “transparent, event-based pricing”
- “startups and technical teams that need high-volume tracking”
- “dominant 'everything store' for product data”
Sample AEO report
AI observability prompts: the new market
Buyer questions in AI observability: tracing agent conversations, tracking token costs, and evaluating LLM outputs in production. Langfuse, Helicone, Braintrust, and Arize are scored here.
10%
Share of voice — AI assistants
1% citation rate across the five chat surfaces
21%
Share of voice — coding agents
2% citation rate across the coding agents
76%
langfuse.com share of voice
Across the same prompts and surfaces
53/64
Cells with zero mentions
Each one is a buyer hearing about someone else
| AI assistants | Coding agents | |||||||
|---|---|---|---|---|---|---|---|---|
| Prompt | ChatGPT | Gemini | Claude | Perplexity | AI Overview | Claude Code | Codex | Cursor |
| What are the best LLM observability tools in 2026?product recommendation | ||||||||
| What should I use to track token costs and latency for my AI product?product recommendation | ||||||||
| What is a good open-source tool for tracing and debugging LLM applications?product recommendation | ||||||||
| What are good alternatives to Langfuse for LLM analytics?comparison | ||||||||
| How do the leading LLM evaluation and observability platforms compare?comparison | ||||||||
| How do I trace multi-step AI agent conversations to find where they go wrong?problem solving | ||||||||
| How can I connect LLM usage data to product analytics to see which AI features drive retention?problem solving | ||||||||
| Who are the leading vendors in AI observability and evaluation?industry expert | ||||||||
- Primary recommendation
- Mentioned secondarily
- Not mentioned
- No data
Click any dot with a run count to inspect the captured AI answers, cited URLs, and highlighted mentions behind that result.
Who buyers hear about instead
Every company the captured answers named, ranked by share of voice (answers that mention them / answers captured).
- Langfuselangfuse.com78%93/120
- Arize Phoenixarize.com62%74/120
- Braintrustbraintrust.dev48%57/120
- Heliconehelicone.ai46%55/120
- LangSmithlangchain.com37%44/120
- LangSmithsmith.langchain.com26%31/120
- Datadogdatadoghq.com22%26/120
- OpenTelemetryopentelemetry.io18%22/120
- Cometcomet.com18%21/120
- Portkeyportkey.ai18%21/120
What gets cited
Domains behind the URLs cited in these answers (598 citations captured). Click a domain to inspect the URLs behind it.
| Domain | Citations |
|---|---|
| 15 | |
| 46 | |
| 45 | |
| 41 | |
| 37 | |
| 24 | |
| 17 | |
| 16 | |
| 16 | |
| 15 |
- Citations
- 15
- Citations
- 46
- Citations
- 45
- Citations
- 41
- Citations
- 37
- Citations
- 24
- Citations
- 17
- Citations
- 16
- Citations
- 16
- Citations
- 15
Prompts where PostHog is absent
Zero mentions in any captured answer, on any surface. Each one is a content target — click a prompt to see who wins the answer and what gets cited.
Where the citations come from
Every cited URL classified: the client’s own pages, competitor pages, social and community threads, institutional references, or earned media. Each surface leans differently — that changes the playbook per platform.
| Surface | Owned | Competitor | Social & community | Institutional | Earned |
|---|---|---|---|---|---|
| ChatGPT | — | 47 | 2 | — | 29 |
| Gemini | 2 | 13 | 10 | — | 67 |
| Claude | 6 | 14 | 5 | 3 | 100 |
| Perplexity | 3 | 13 | 8 | — | 104 |
| AI Overview | — | 8 | 6 | — | 27 |
| Codex | — | 43 | 7 | — | 33 |
| Cursor | 4 | 20 | 5 | — | 19 |
| All surfaces | 15 | 158 | 43 | 3 | 379 |
How the answers talk about PostHog
Across 17 answers that mention PostHog: 16 positive, 1 neutral, 0 negative.
Phrases the answers attach to PostHog
- “PostHog AI Observability”×2
- “PostHog LLM Analytics”×2
- “strong default”
- “LLM metrics tied to product analytics”
- “users, features, funnels, alerts”
- “one place”
- “best default”
- “costs roll up by user, model, feature, trace, or custom property”
- “LLM data sits next to conversion, retention, and feature usage”
- “not just a provider dashboard total”
- “Product + LLM observability together”
- “Traces generations alongside product analytics, funnels, and user context”
- “LLM traces tied to real user behavior in production”
- “AI observability tied to product analytics”
- “Connects AI traces with product analytics, session replay, experiments, and feature flags”
- “Less specialized as an LLM-only observability tool”
- “product-analytics-adjacent platforms”
- “Teams that already use PostHog (or want LLM quality tied to product analytics)”
- “LLM traces linked to users, sessions, feature flags, experiments”
- “Not a standalone "eval lab" like Braintrust”
Technical AEO
Can the AI systems even read PostHog?
Visibility starts with access. The bots behind AI answers must be allowed in. And the content must be readable without JavaScript.
AI crawler access
Whether posthog.comlets each AI bot in — from robots.txt, plus a live fetch with each search bot’s user agent to catch firewall-level blocks robots.txt doesn’t show. Training bots collect data for future models. Search bots power live AI answers — blocking those removes you from answers. User-fetch agents load a page when a person asks about it.
| Bot | Type | robots.txt | Live fetch |
|---|---|---|---|
| GPTBot | Training | Allowed | — |
| OAI-SearchBot | Search | Allowed | OK |
| ChatGPT-User | User fetch | Allowed | OK |
| ClaudeBot | Training | Allowed | — |
| Claude-User | User fetch | Allowed | OK |
| Claude-SearchBot | Search | Allowed | OK |
| PerplexityBot | Search | Allowed | OK |
| Perplexity-User | User fetch | Allowed | OK |
| Google-Extended | Training | Allowed | — |
| meta-externalagent | Training | Allowed | — |
| Bytespider | Training | Allowed | — |
| CCBot | Training | Allowed | — |
Content without JavaScript
Most AI retrieval bots don’t run JavaScript. For each key page: how much of the fully rendered text is already in the raw HTML. Flagged pages hide substance from the bots that write the answers.
| Page | Visible without JS | Verdict |
|---|---|---|
| https://posthog.com/ | 88% | OK |
| https://posthog.com/pricing | 100% | OK |
| https://posthog.com/docs | 90% | OK |
| https://posthog.com/ai-observability | 99% | OK |
Hygiene & agent readiness
- Presentllms.txt — low-cost hygiene; whether AI systems consume it is still unproven, so we report it honestly as a nice-to-have.
- Missingllms-full.txt — same honest framing as llms.txt.
- AvailableMCP server — lets coding agents use the product directly.
- AvailableLLM-oriented docs — docs formatted for AI consumption (checked by a person, not crawled).
Earned media
The sources the answers trust
The answers cite these sites on prompts that name a competitor but not PostHog. These are the places to earn a mention. People decide which ones to pursue; the data only points.
01
confident-ai.com
Cited 23 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What are good alternatives to Langfuse for LLM analytics?”
- “How do the leading LLM evaluation and observability platforms compare?”
+1 more promptsShow fewer
- “Who are the leading vendors in AI observability and evaluation?”
02
langchain.com
Cited 23 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What should I use to track token costs and latency for my AI product?”
- “What is a good open-source tool for tracing and debugging LLM applications?”
+4 more promptsShow fewer
- “What are good alternatives to Langfuse for LLM analytics?”
- “How do the leading LLM evaluation and observability platforms compare?”
- “How do I trace multi-step AI agent conversations to find where they go wrong?”
- “Who are the leading vendors in AI observability and evaluation?”
03
openobserve.ai
Cited 17 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What is a good open-source tool for tracing and debugging LLM applications?”
- “What are good alternatives to Langfuse for LLM analytics?”
+1 more promptsShow fewer
- “How do the leading LLM evaluation and observability platforms compare?”
04
getmaxim.ai
Cited 13 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What should I use to track token costs and latency for my AI product?”
- “How do I trace multi-step AI agent conversations to find where they go wrong?”
05
mlflow.org
Cited 12 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What should I use to track token costs and latency for my AI product?”
- “What is a good open-source tool for tracing and debugging LLM applications?”
+1 more promptsShow fewer
- “How do the leading LLM evaluation and observability platforms compare?”
06
firecrawl.dev
Cited 10 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What is a good open-source tool for tracing and debugging LLM applications?”
- “How do the leading LLM evaluation and observability platforms compare?”
07
galileo.ai
Cited 10 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What is a good open-source tool for tracing and debugging LLM applications?”
- “How do the leading LLM evaluation and observability platforms compare?”
+2 more promptsShow fewer
- “How do I trace multi-step AI agent conversations to find where they go wrong?”
- “Who are the leading vendors in AI observability and evaluation?”
08
kosmoy.com
Cited 10 times in answers without PostHog.
- “What are good alternatives to Langfuse for LLM analytics?”
- “How do the leading LLM evaluation and observability platforms compare?”
- “Who are the leading vendors in AI observability and evaluation?”
09
docs.langchain.com
Cited 9 times in answers without PostHog.
- “What are the best LLM observability tools in 2026?”
- “What are good alternatives to Langfuse for LLM analytics?”
- “How do the leading LLM evaluation and observability platforms compare?”
Baseline measured August 22, 2026: 16 prompts, 240 captured answers across ChatGPT, Gemini, Claude, Perplexity, and Google AI Overview, plus the coding agents Claude Code, Codex and Cursor. This is the AEO baseline; re-measurement is quarterly.
Chat surfaces are asked with native web search turned on where the provider supports it, plus direct Google AI Overview data. Each prompt is sampled more than once. The three coding-agent surfaces run the real products: Claude Code, the Codex CLI, and Cursor, headless, with web tools on and no repo context. That is what a developer actually experiences. Agents often answer from memory without searching. Each captured answer records whether it searched, and the report says so. The headline scorecards cover the five chat surfaces. The coding agents get their own scorecards.
Every number on this page keeps its raw model answers behind it. Every probe can be checked. This is an independent analysis by Literally.dev.
Your turn
Want this for your company?
Request the free analysis. We run the same measurement on your brand: eight AI tools, coding agents included, every answer saved.