Open source · five-pillar audit · AGPL-3.0

Get found. Get cited.
Get chosen — by the agents that act.

Seryvon is the only open-source audit that scores your site across all five layers of modern discovery — SEO, GEO, GSO, AEO and ASO — with transparent, deterministic, fully traceable math.

$ uvx --from seryvon seryvon audit yoursite.com
AGPL-3.0·BYOK·runs 100% free
Real Seryvon Docker audit overview in English with a 94 out of 100 illustrative score
Real Docker dashboard capture · illustrative audit data
691
tests passing
100%
test coverage
0
mypy strict errors
<2%
variance on re-run
v1.0.0rc4
current release
5
pillars unified
Why now

The tools you trust are blind to the new game.

In 2026, AI agent traffic to e-commerce sites grew by several hundred percent in a single year. Autonomous agents now book, compare, select and purchase — without ever asking a human to click. Ahrefs tracks your Domain Rating. Semrush monitors your keyword rankings. Neither can tell you whether ChatGPT cites you, whether Perplexity recommends you — or whether an autonomous agent can find, parse and choose you.

That gap has a name: Signal clarity. The information overload of the AI era makes it harder for the right signal to reach the right decision-maker — human or machine. Seryvon is the audit built to close that gap.

Blind spot 01
Cost and opacity

Ahrefs, Semrush and Moz charge $100–500+/month per seat and deliver scores you cannot audit, reproduce or customize — Domain Rating, Authority Score. Pure black boxes.

Blind spot 02
Generative coverage

GEO and AEO are afterthoughts in commercial tools — added late to premium tiers, methodology proprietary, scores not reproducible across runs.

Blind spot 03
The agentic blind spot

No commercial tool audits agentic readiness. Whether a shopping agent or booking agent can discover and choose you is simply unmeasured — anywhere, by anyone else.

Three eras of discovery

SEO got you ranked. GEO gets you cited. ASO gets you chosen by the agents that act.

2000 → 2022
Be Ranked
SEO

Humans type queries. Engines rank pages. Winning meant high positions, fast load times, clean metadata and authority links.

→
2022 → 2024
Be Cited
GEO · AEO · GSO

AI models answer queries by synthesising the web. Winning means being the source a model quotes — entity clarity, E-E-A-T, structured Q&A.

→
2025 →
NOW
Be Chosen
ASO — Agentic Search Optimization

Autonomous agents compare, book, buy and select on behalf of users. Winning means exposing machine-readable actions, MCP endpoints and a coherent cross-surface identity.

Five pillars, one engine

Every layer of modern discovery, measured and unified.

One deterministic engine scores all five pillars and rolls them into a single weighted, fully traceable result. Every criterion links to a measured value.

SEO
Search Engine Optimization

Classic organic compliance — crawlability, metadata, Core Web Vitals, maillage interne, authority. 28 criteria measured.

weight0.30
GEO
Generative Engine Optimization

Machine retrievability, entity density, E-E-A-T signals and real LLM citation tracking (BYOK). 13 criteria.

weight0.22
GSO
Google AI Overviews

Rich-snippet eligibility, extractable Q&A and measured presence in AI Overviews via SERP API (BYOK). 8 criteria.

weight0.18
AEO
Answer Engine Optimization

Source credibility, atomic answers, Knowledge Graph presence (Wikidata) and verifiable dating. 8 criteria.

weight0.15
ASO NEW
Agentic Search Optimization

Can autonomous agents discover, parse and choose you? MCP readiness, action schema, AI endpoints, NLWeb. 9 criteria.

weight0.15
ASO v2 v2 roadmap
Dynamic Agent Testing

Live testing: does a real agent comparator select you? Mechanics close to citation tracking (BYOK, K repetitions). Backlog.

weight—
The differentiator — ASO

The only open-source audit that scores whether AI agents can choose you.

No competitor scores this. GEO Optimizer includes a WebMCP check but it's an unscored bonus. Ahrefs, Semrush and Moz have no ASO signal at all. Seryvon makes it a full scored pillar — 9 deterministic criteria, all measurable without calling an agent.

GEO vs ASO — the key distinction
GEO/AEO/GSO → AI answers a human by citing you — you get visibility.
ASO → An agent that acts (books, buys, compares) can find and choose you — you get business.
mcp_readiness w:1.8
MCP Readiness

WebMCP API detected: navigator.modelContext.registerTool() or toolname/tooldescription HTML attributes.

potential_actions w:1.6
potentialAction

JSON-LD potentialAction: SearchAction, BuyAction, OrderAction, ReserveAction.

action_schema w:1.5
Rich action schemas

Product+Offer (price+availability), Service+areaServed, Event+dates, HowTo+steps.

ai_discovery w:1.4
AI discovery endpoints

/.well-known/ai.txt, /ai/summary.json, /ai/faq.json, /ai/service.json validated (HTTP 200 + JSON).

brand_coherence w:1.3
Brand coherence

Consistent identity across site, Wikidata, LinkedIn, Crunchbase, G2 — what an agent sees when it resolves you.

agent_access w:1.2
Agent bot access

robots.txt and CDN allow OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended and other agent crawlers.

nlweb w:1.0
NLWeb readiness

NLWeb endpoint conformance — the emerging standard for agent interrogation of websites.

accessible_forms w:0.8
Agent-usable forms

Inputs labelled with label/for, aria-label or descriptive placeholder + defined action and method.

openapi w:0.8
OpenAPI exposed

Links to /api-docs, /swagger or openapi.json/yaml — consumable directly by agent tooling.

Seryvon Docker agent readiness and ASO signals
Agent readiness: ASO signals and declared actions
How Seryvon compares

Built to fill every blind spot the incumbents left open.

OpenSEO (MIT, ~2.1k★) excels at classic SEO with BYOK DataForSEO but has no transparent scoring and no ASO. GEO Optimizer (MIT, ~471★, 71 releases) leads on GEO/AEO but covers ASO only as an unscored bonus check, and does not do full SEO. The commercial trio (Ahrefs, Semrush, Moz) charge $100–500+/month for proprietary black-box scores — and none audit agentic readiness.

Feature
Ahrefs
Semrush
Moz
OpenSEO
GEO Opt.
Seryvon
Full SEO audit
✓
✓
✓
✓
✗
✓
Transparent deterministic scoring
✗
✗
✗
✗
✓
✓
GEO / LLM citation tracking
✗
◐ add-on
✗
◐ partial
✓
✓
GSO / AI Overviews
✗
◐ add-on
✗
✗
✓
✓
AEO / answer engines
✗
✗
✗
✗
✓
✓
ASO / agentic (scored)
✗
✗
✗
✗
◦ unscored
✓
Open source
✗
✗
✗
MIT
MIT
AGPL-3.0
Free / BYOK
✗
✗
✗
✓
✓
✓
CI/CD integration
✗
✗
✗
✓
✗
✓
Starting price
$99+
$130+
$99+
free
free
free

★ GEO Optimizer includes WebMCP as an unscored "bonus check", not a scored pillar. † OpenSEO covers basic GEO signals without LLM citation tracking or deterministic scoring.

Seryvon Docker citation tracking dashboard by engine
Citation Tracking: observe visibility across AI engines
How it works

From crawl to a score you can defend.

01
Crawl

Async BFS crawl that respects robots.txt, rate-limits politely, renders JS via Playwright where needed, bounded by depth and page count you configure.

02
Measure

Collects all raw signals: HTTP, JSON-LD, Core Web Vitals (PSI), LLM citations, SERP presence, MCP readiness, AI discovery endpoints, Wikidata.

03
Score

Deterministic 0–100 per criterion — pure functions, no randomness. Criteria are weighted into five pillars and one global score. Two runs → same result (variance < 2%).

04
Export & trace

JSON, HTML, PDF or Markdown. Every score links to its raw measured value, thresholds and formula. The action plan is auto-prioritised by impact × severity ÷ effort.

Built to production standards

Open source does not mean prototype.

Seryvon v1.0.0rc4 ships with the discipline of a production service: mypy strict, ruff, 100% coverage, deterministic scoring tests and SSRF guard on every network entry point.

seryvon audit --stats
✓ 691 tests passed
✓ mypy strict — 0 errors
✓ ruff — clean
→ coverage: 100%
→ variance on re-run: <2%
691
tests passing
100%
test coverage
0
mypy strict errors
<2%
score variance on re-run
~4min
per 100-page audit
SSRF
guard on all inputs
Roadmap

Seryvon Core v1.0.0rc4 is public.

The v1.0.0rc4 open-core release is publicly tagged with five pillars, React UI, i18n, BYOK and SSRF guard. Hosted monitoring remains separate and forthcoming.

✓ v0.1 SEO audit, CLI, JSON/HTML export
✓ v0.2 Five pillars on-page — ASO static included, 4 export formats
✓ v0.3 LLM citation tracking multi-engine (BYOK)
✓ v0.4 Measurement profile + comparison M6
✓ v1.0.0rc4 React UI, EN/FR, full BYOK, SSRF guard — 691 tests, 100% coverage current
○ v1.0 v1.0.0rc4 public OSS release — stable hardening next
○ Cloud Operated offering — monitoring, history, multi-user, white-label
Seryvon Docker BYOK encrypted connector configuration
BYOK: connect your own providers with encrypted key storage
Open core

Free tells you where you stand. Paid watches, remembers and collaborates.

The frontier is never a pillar — all five pillars are free forever. The frontier is a one-off audit versus a platform operated over time. You only pay for what costs real infrastructure.

Community Free
Self-hosted · AGPL-3.0
✓All five pillars — SEO, GEO, GSO, AEO & ASO
✓Transparent deterministic scoring
✓BYOK for every API (LLM, SERP, PSI, GSC)
✓CLI, exports (JSON / HTML / PDF / MD), Docker
✓Rank tracking via Search Console (free)
✓Local audit history
✓MCP server — audit from Claude or Cursor
Available now
Pro / Team Operated
Managed hosting · continuous
✓Everything in Community
✓Continuous monitoring of all five scores
✓Alerting when a pillar drops below your threshold
✓Scheduled rank & LLM citation tracking
✓Long-horizon history & Share of Voice
✓Multi-user, roles, shared workspaces
✓White-label client reports
✓Managed hosting — no Docker to run
Join the waitlist
Enterprise — SSO, SLA, dedicated support and assisted deployment, on request.
Seryvon Docker audit history and score comparison
History and score comparisons
Seryvon Docker rank tracking via Search Console
Rank tracking via Search Console
Zero-friction adoption

Drop it into the workflow you already have.

01
Zero-install CLI

uvx --from seryvon seryvon audit <url> — try it with nothing installed. Running in under 30 seconds.

02
MCP server

Audit straight from Claude or Cursor — Seryvon exposes its own MCP server and eats its own dog food.

03
CI/CD gate

A GitHub Action that fails the build when your ASO or GEO score drops below a threshold you define.

04
ASO Ready badge

An embeddable agentic-readiness badge for your site — earns backlinks, demonstrates credibility.

Contact

Star it, try it, or reach out.

Seryvon v1.0.0rc4 is public and ready for self-hosted evaluation. Read the repository, run an audit, or write about Enterprise and partnerships.

Join the waitlist

Enter your email and we'll reach out when the managed hosting offer goes live.

This landing does not create an account and does not store your email locally. Clicking the button opens your email client so you stay in control of the send action. Read the privacy policy.

Privacy by design

A lightweight landing, with clear data handling.

This page is designed to limit personal-data collection to what is strictly necessary for early-contact requests. It uses explicit consent for the waitlist and keeps legal information visible from the footer.

What is Seryvon?

Seryvon is an open-source web visibility audit tool for teams that need evidence about how a website can be found, understood, cited and chosen. It measures five distinct pillars: SEO, GEO, GSO, AEO and ASO. The score describes signals available in the inspected run; it is not proof of ranking, inclusion in an AI answer or successful agent activity. Read the measurement methodology and limitations before interpreting results.

  • SEO: can a search engine discover and interpret the site?
  • GEO: can a generative engine retrieve and cite the content?
  • GSO: is the site prepared for Google AI search experiences?
  • AEO: can an answer engine extract a direct, supported answer?
  • ASO: can an autonomous agent discover and understand declared actions?

How does an AI visibility audit work?

Seryvon crawls the selected website, records technical and editorial signals, and combines them with optional provider measurements. The report identifies the URL, scope, date, coverage and rule evidence. A readiness result describes what the inspected page exposes; it does not claim that Google, ChatGPT or Perplexity has cited the page.

  • Record crawl and provider configuration.
  • Separate measured readiness from observed citations.

Does a Seryvon score prove an AI citation?

No. A score and an observed citation are separate findings. A citation requires a recorded response from a specific engine, with its prompt, model, retrieval mode and collection date.

  • Readiness describes page signals.
  • Citations require engine-specific evidence.

Is Seryvon available as a hosted service?

The open-core implementation and documentation are public. The managed hosted product is coming soon. Visitors can start with the quickstart, inspect the methodology and review the repository.

  • Open core: public source and local evaluation.
  • Managed service: planned, not yet generally available.