VE LAB · Internal Tooling Evaluation

GEO & Agentic-SEO Tooling: Benchmark & Recommendation

Scored for the ve-lab.exe.xyz → velab.org deployment (static site, Caddy, Claude Code workflow). Compiled 2026-06-30.

Eight open-source GEO / agentic-SEO tools were reviewed, scored on a 5-criterion balanced benchmark, and cross-checked against 30 days of live GitHub and social signal. One tool wins decisively; one late entrant earns a second look.

★ Recommendation — 100 / 100

Auriti-Labs/geo-optimizer-skill

The only candidate that covers both halves of the velab.org need in one tool: static-site auditing (robots.txt, llms.txt, JSON-LD, citability) and verification of whether ChatGPT / Perplexity / Gemini actually cite the site. MIT-licensed, Python-only (no Node, no GPU), ships an MCP server for native Claude Code use, and the CLI is free forever. Live GitHub data shows same-day commit activity — actively maintained, not rotting.

Use it to audit ~/Documents/velab-site/public/ before the DNS cutover, then schedule geo citations for ongoing monitoring against Ollama Cloud / Perplexity.

The ranking

Five criteria, equal weight (20% each), each scored 0–5. Self-host scoring is tuned to VE LAB's posture: a bring-your-own LLM key is a minor cost (Ollama/Perplexity already in use), while vendor lock-in and GPU dependence are near-disqualifying.

#	Tool	INT	SSF	CIT	SH	MH	Total
1	Auriti-Labs/geo-optimizer-skill	5	5	5	5	5	100
2	onvoyage-ai/voyage-geo-agent	5	0	5	4	2	64
3	aaron-he-zhu/seo-geo-claude-skills	5	3	0	5	3	64
4	IJONIS/geo-lint	5	3	0	5	2	60
5	ngstcf/ai-seo-auditor	3	5	0	4	2	56
6	github/awesome-copilot (SEO-AI agent)	2	3	0	5	4	56
7	AI2HU/gego	3	0	5	2	4	56
8	cxcscmu/AutoGEO	3	0	5	0.5	2	42
9	ai-search-guru/getcito	1	0	3	0.5	4	34

INT = Claude/agent integration · SSF = static-site GEO fit · CIT = citation tracking · SH = self-host / cost / no lock-in · MH = maintenance health.

Why the spread looks like this

One axis can't carry a tool

Under equal weights, single-purpose tools (ai-seo-auditor nails auditing; voyage/gego nail citation tracking) all cluster at 52–64 because a zero on any criterion caps them. Auriti wins by being strong everywhere.

Your posture moved the mid-field

voyage-geo-agent jumped to #2 once "BYO paid key" became a minor penalty — you already pay for inference. gego, getcito and AutoGEO sank on heavy infra, Firebase/Azure lock-in, and GPU requirements.

30-day market signal

A /last30days run in GitHub project-mode pulled live repo health and social chatter for all seven shipping tools. Findings:

Benchmark validated. Live star order matched the scoring; Auriti, voyage, AutoGEO, getcito and ai-seo-auditor all showed same-day (2026-06-30) activity.
Research is hardening. Tooling now cites real papers — Princeton KDD 2024, AutoGEO (ICLR'26, a CMU+Vody RL method ~140× cheaper than prompting GPT-4, +36% visibility), and a new AgenticGEO self-evolving-system paper. Auriti's edge is that it operationalized the papers that AutoGEO leaves as research code.
Named-tool social buzz is near-zero. These are early-adopter dev tools — judge by stars + commit recency, not mentions.

Late entrant, corrected on review: GEO-Lint (#4, 60)

IJONIS/geo-lint wasn't in the original candidate set — a 30-day scan caught it on X and twice on Hacker News. It's an MIT, Node-only linter built for an agent to "run, read, fix, and re-lint automatically." A feasibility review then corrected two facts: (1) the npm package is @ijonis/geo-lint@0.2.1 — not the unscoped geo-lint, which is an unrelated GPS library; and (2) it audits .md/.mdx content source, not built HTML, so for a static-HTML target like velab.org it needs a custom adapter. That dropped its static-site-fit score from 5 to 3, moving it from a provisional #2 to #4 (60), behind voyage and seo-geo-claude-skills. Still useful where markdown source exists — but not the drop-in HTML auditor it first appeared to be.