Building jev-seo: Zero-Cost SEO and GEO Search Radar in Rust

Traditional SEO tooling is broken for independent developers. Semrush and Ahrefs cost upwards of $130 per month. Open-source alternatives still require paid DataForSEO API keys or heavy browser runtimes.

At the same time, search engine mechanics have bifurcated:

1. Traditional SERP ranking: Crawlers index static metadata, Schema.org entities, and semantic content density.
2. Generative Engine Optimization (GEO): AI answer engines (Perplexity, ChatGPT Search, Claude) synthesize answers from top snippets and cite concise, factual paragraph blocks.

I built jev-seo to give developers and coding agents a complete SEO and GEO audit engine with zero subscription costs.

+-------------------+        +----------------------+        +---------------------+
| Target Site / URL |        | jev-seo (Rust CLI)   |        | TypeSafe AI (Jev)   |
| Markdown / HTML   | -----> | Local DuckDuckGo SERP| <----> | /v1/systemone       |
+-------------------+        | Fast HTML Parser     |        | 70-120ms semantic   |
                             +----------------------+        +---------------------+
                                        |
                                        +---> .jev-seo.db (Local SQLite)
                                        +---> Terminal Tables & JSON-RPC MCP

Zero-Cost Architecture

jev-seo runs as a single static Rust binary with no headless Chrome dependencies.

• Zero-API SERP scraping: Queries DuckDuckGo HTML and autocomplete endpoints directly. No paid proxies or credit card registrations required.
• Fast static audits: Scans hundreds of local HTML and Markdown files in under 400ms using fast-html-parser and gray_matter.
• System One semantic scoring: Uses TypeSafe AI Jev (jev-latest) to score search intent alignment, query relevance, and citation viability at 70ms latency.
• Local SQLite telemetry: Stores historical rankings and audit runs in a local SQLite file (.jev-seo.db) with zero external network tracking.

src/
  main.rs          # Clap CLI parsing and subcommands
  mcp.rs           # Stdio JSON-RPC 2.0 MCP server for agent integration
  audit.rs         # Local file scanner: titles, meta tags, and open graph
  schema.rs        # JSON-LD 2026 validator (SoftwareApplication, Article)
  robots.rs        # Robots.txt parser with AI crawler radar
  serp.rs          # DuckDuckGo HTML search scraper and parser
  brief.rs         # SERP content brief synthesizer
  geo.rs           # Generative Engine Optimization scorer
  db.rs            # SQLite storage with rusqlite

Subcommands and Capabilities

The CLI provides eight specialized subcommands:

# 1. Audit local project files for SEO defects
jev-seo audit ./content

# 2. Check JSON-LD Schema.org validity
jev-seo schema https://example.com

# 3. Inspect robots.txt rules for AI crawlers (GPTBot, ClaudeBot)
jev-seo robots https://example.com/robots.txt

# 4. Synthesize a content brief from live SERP competition
jev-seo brief "offline expense tracker android"

# 5. Run Generative Engine Optimization citation evaluation
jev-seo geo "What is the fastest grep tool for AI agents?"

# 6. Query live search rankings and competitor positions
jev-seo query "driverless cdp browser"

# 7. Track keyword positions over time
jev-seo track "agentic coding stack" --domain example.com

# 8. Start the MCP server for AI coding agents
jev-seo --mcp

GEO Scoring and Citation Blocks

Generative search engines do not read 3,000-word articles end to end. They extract grounded factual segments between 134 and 167 words that directly answer specific queries.

jev-seo geo runs input text through TypeSafe AI System One to evaluate two distinct metrics:

1. Information density: Does the paragraph answer the user prompt immediately without introductory fluff?
2. Factual claim clarity: Are architectural claims backed by concrete metrics (latency, binary size, memory usage)?

Evaluation:
  Query: "fastest rust grep for coding agents"
  Citation readiness: 94/100
  Word count: 148 words (optimal band: 134-167)
  Verdict: PASSED

Agent Integration via MCP

Coding assistants like Claude Code and Antigravity can run jev-seo as an stdio MCP server. When an agent writes documentation or landing pages, it queries jev-seo to verify Schema.org syntax and check keyword visibility before committing.

The binary compiles to 3.8MB on Windows and Linux with zero background daemon memory.