Yes. FasterGEO is Apache-2.0 open source and self-hosted. You bring your own AI engine API keys, or use manual sampling sheets with zero keys. Your data never leaves your machine.

Do I need API keys to use FasterGEO?

No. Site auditing needs no keys at all. Answer sampling works with any keys you have, and engines without keys fall back to manual sampling sheets that import into the same metrics pipeline.

Why should I trust FasterGEO's numbers?

Every metric is publicly defined and traceable to the code that computes it, in the open METHODOLOGY document. Unmeasured values render as unmeasured, never as zero. Verdicts ship with verbatim quotes, and statistics on the site are footnoted to their sources.

How is FasterGEO different from AI visibility monitoring tools?

Three ways. It measures the head of the funnel — whether AI even knows your brand or confuses it with another company — with quoted evidence, not just mention counts. It closes the loop: findings become tickets with machine-verifiable acceptance criteria, re-checked automatically. And it is the only tool covering both China and global engines.

Open-source GEO platform · 18 AI engines, China + global · Apache-2.0 · v0.10

AI is already answering questions about your brand.
Read the evidence.

Q: What is GEO (Generative Engine Optimization)?

GEO is a practice of making AI engines — ChatGPT, Claude, Perplexity, Doubao, DeepSeek — mention and cite your brand when users ask them for recommendations. Unlike SEO, which optimizes for ranked links, GEO optimizes for being part of the answer itself.

Q: Which AI engines does FasterGEO cover?

18 engines across two markets: ChatGPT, Claude, Gemini, Grok and Perplexity globally; GLM, Doubao, DeepSeek, Kimi, MiniMax, Qwen, ERNIE and Spark in China via API — plus manual sampling sheets for engines without public APIs, such as Baidu AI, Nano AI, and Google AI Overviews.

Your next customer just asked ChatGPT — or Doubao — which brand to pick. FasterGEO shows you the answer they got, then turns every gap into a fix a machine verifies — with the verbatim quote attached.

$ npx fastergeo bootstrap --root yoursite.com copy Read the methodology →

Free · self-hosted · your data stays on your disk

Exhibit A · verbatimrecognition judge — first field run

“…mainly known for aftermarket car exterior parts — bumpers, grilles, interior trim…”

A real AI answer describing a custom-merch platform — caught, quoted, filed as a P0. Evidence is never paraphrased.

First check

What do AI crawlers see on your site?

The scan is the product, not a teaser — the same six-dimension audit the CLI runs, on one page of yours, free, no sign-up. Two failure modes dominate the sites we've scanned: pages that render empty to non-JavaScript crawlers, and robots rules that silently ban the wrong bots.

Fetches raw HTML exactly as AI crawlers do — no JavaScript executed. That's the point.

Brand owner → this scan · Developer → npx fastergeo · Agency → the 90-day playbook

The first case file

We ran the first scan on our own site.
The machine found what humans had missed.

Twenty minutes, three findings. Every number below is from that real first diagnosis — nothing composited, nothing invented.

23visible words

Product pages were empty shells

429 KB of HTML, 23 visible words. AI crawlers don't execute JavaScript — every product page was literally telling AI “No products for this category.”

2engines confused

Mistaken for entirely different companies

One engine described the brand as an auto-parts maker (Exhibit A, above). Another confidently fabricated user reviews for a same-named menswear tailor. Not an outlier: ChatGPT Search misattributes sources in 76.5% of tested answers1 — and flags uncertainty in almost none of them.

0%unprompted mentions

Absent from every recommendation

Across four engines and 64 buying-intent questions, competitors owned every answer. Most brands aren't losing the AI recommendation race — they were never in it.

The Brand Entity Funnel

Dashboards measure the tail.
Most brands break at the head.

Monitoring tools count mentions. But five questions decide whether AI sells for you or against you — and for most brands, the answer breaks at question one. FasterGEO measures all five, with a judge that must quote its evidence.

✕ broke here1 · Does AI know you?Probe questions, LLM-judged — never name-echo counting

✕ broke here2 · Does it confuse you?Mistaken identity = P0 alert with the quote attached

measured3 · Does it mention you?Unprompted mention rate, with 95% intervals

measured4 · Does it rank you?Your position when buyers ask "which one?"

measured5 · Does it cite you?Your domain in the sources — plus whose domains AI does trust

✕ marks where our own first scan broke. Where does yours?

The loop

Not a dashboard. A closed loop — watch it run.

Captured output from the real first project. Diagnosis becomes tickets; tickets are verified by re-crawls, not verbal confirmations.

Captured terminal outputcustyle · first period

A ticket flips to done only when a re-crawl proves it — and back to regressed when the fix rots.

Against the field

Monitoring is becoming free.
The loop is the product.

Google Search Console and Bing Webmaster Tools now publish official AI-performance reports; HubSpot sells monitoring for $50 a month.6 Counting mentions is a commodity on an 18-month clock. What survives: knowing why you lose, fixing it, and proving the fix — in the open.

	Monitoring SaaS	FasterGEO
China engines	not covered	10 engines — Doubao, DeepSeek, GLM, Kimi, Qwen, ERNIE… API + manual sheets
Recognition	name-echo counting	knows / unknown / confused — with the verbatim quote, or the verdict stays unverified
Action loop	suggestions	tickets → re-crawl → machine-verified done, auto-reopened on regression
Content	free-form drafts	every number must trace to a sourced fact — drafts that can't, fail the gate
Metrics	black-box score	every formula public and code-traceable; unmeasured stays unmeasured, never zero
Your data	vendor cloud	your disk
Price	$99–5,000/mo	$0 — Apache-2.0, bring your own keys or none

Migrating from GeoLook? fastergeo metrics --format geolook re-scores your existing samples directly.

The register

Eighteen engines. Two markets.
Never averaged together.

China · API

GLM 智谱 · Doubao 豆包 · DeepSeek · Kimi · MiniMax · Qwen 通义 · ERNIE 文心 · Spark 星火

China · sheets

Baidu AI 百度 · Nano AI 纳米

Global · API

ChatGPT · Claude · Gemini · Grok · Perplexity

Global · sheets

ChatGPT web · Claude web · Google AI Overviews

fastergeo check diagnoses every key — no key / auth failed / authenticated-but-model-not-enabled — each with an actionable hint. HTTPS_PROXY honored.

Notes & sources

The numbers behind the shift, footnoted.

84% of AI citations come from earned media, not brand sites2 — the third-party domains AI already cites are your target list, and FasterGEO extracts them from your own samples. Brands cited inside AI Overviews see +35% organic CTR (+91% paid)3. Meanwhile 41% of sites block GPTBot without knowing it4, and only 14% of marketers measure the AI channel at all5. The window where measuring is an advantage is open. It won't stay open.

76.5% misattribution — Columbia Tow Center testing of ChatGPT Search. ziptie.dev
84% earned media — Muck Rack, 25M citations analyzed. machinerelations.ai
+35% / +91% CTR when cited — AirOps. airops.com
41% unknowing blocks — Digital Strategy Force crawler audit. digitalstrategyforce.com
14% measuring — AIThinkerLab, 2026. aithinkerlab.com
Official AI reports — Google Search Console Gen-AI performance (developers.google.com), Bing Webmaster Tools AI Performance; HubSpot AEO pricing (hubspot.com).

Questions

Asked often, answered plainly.

What is GEO (Generative Engine Optimization)?

The practice of making AI engines — ChatGPT, Claude, Perplexity, Doubao, DeepSeek — mention and cite your brand when users ask for recommendations. SEO optimizes for ranked links; GEO optimizes for being part of the answer itself.

Is FasterGEO free?

Yes. Apache-2.0, self-hosted. Bring your own engine API keys, or use manual sampling sheets with zero keys. Your data never leaves your machine.

Which engines are covered?

Eighteen across two markets: ChatGPT, Claude, Gemini, Grok and Perplexity globally; GLM, Doubao, DeepSeek, Kimi, MiniMax, Qwen, ERNIE and Spark in China via API — plus manual sheets for engines without public APIs, such as Baidu AI and Google AI Overviews.

Do I need API keys?

No. Site auditing needs none at all. Sampling works with any keys you have; engines without keys fall back to manual sheets that import into the same metrics pipeline.

Why should I trust these numbers?

Because you don't have to take them on faith. Every metric is publicly defined and traceable to the code that computes it — see the methodology. Unmeasured values render as unmeasured, never as zero. Verdicts ship with verbatim quotes. And every statistic on this page is footnoted to its source — the same fabrication gate the product enforces on your content applies to our marketing.

How is this different from AI visibility monitoring tools?

Three ways. It measures the head of the funnel — whether AI even knows your brand or confuses it with another company — with quoted evidence. It closes the loop: findings become tickets with machine-verifiable acceptance, re-checked automatically. And it's the only tool covering both China and global engines.

AI is already answering questions about your brand.
Read the evidence.

What do AI crawlers see on your site?

We ran the first scan on our own site.
The machine found what humans had missed.

Product pages were empty shells

Mistaken for entirely different companies

Absent from every recommendation

Dashboards measure the tail.
Most brands break at the head.

Not a dashboard. A closed loop — watch it run.

Monitoring is becoming free.
The loop is the product.

Eighteen engines. Two markets.
Never averaged together.

Single pages of the engine, free forever.

The numbers behind the shift, footnoted.

Asked often, answered plainly.

Every day you don't look,
AI is answering without you.

AI is already answering questions about your brand.Read the evidence.

What do AI crawlers see on your site?

We ran the first scan on our own site.The machine found what humans had missed.

Product pages were empty shells

Mistaken for entirely different companies

Absent from every recommendation

Dashboards measure the tail.Most brands break at the head.

Not a dashboard. A closed loop — watch it run.

Monitoring is becoming free.The loop is the product.

Eighteen engines. Two markets.Never averaged together.

Single pages of the engine, free forever.

The numbers behind the shift, footnoted.

Asked often, answered plainly.

Every day you don't look,AI is answering without you.

AI is already answering questions about your brand.
Read the evidence.

We ran the first scan on our own site.
The machine found what humans had missed.

Dashboards measure the tail.
Most brands break at the head.

Monitoring is becoming free.
The loop is the product.

Eighteen engines. Two markets.
Never averaged together.

Every day you don't look,
AI is answering without you.