GEO tools compared: what each kind actually does (including where we fit)

"GEO tools" (generative engine optimization) has become a label covering four different jobs. Vendors blur them because "all-in-one" sells; buyers end up with two tools doing the same thing and a gap where the third should be. Here's the landscape sorted by job — including, transparently, where our own product sits and what it does not do.

Job 1 — Answer tracking: "what do the AIs say about us?"

Tools that run prompts against ChatGPT, Perplexity, Gemini and log whether you're mentioned or cited, over time — rank tracking reborn for answers. Worth paying for when you're monitoring many prompts across many markets; until then, a monthly manual loop in a spreadsheet gets you 80% free, and teaches you what the paid dashboards actually measure. Caveats: answers are stochastic (single runs mislead), and personalization contaminates naive testing.

Job 2 — Content optimization: "will AI cite this page?"

Graders and rewriters that score pages for citability — structure, directness, extractable facts, schema. Useful as a checklist enforcer; the honest limit is that what earns citations is mostly old-fashioned editorial quality, and a score can't write specificity into thin content. If your pages already answer one question each in crawlable HTML, these tools confirm more than they change.

Job 3 — Crawler control: "who may read us?"

The policy layer: robots.txt management, per-bot allow/block, sometimes CDN-level enforcement. The decisions matter more than the tooling — blocking AI search de-lists you, blocking training costs nothing, the split matters. For most sites this genuinely doesn't need a paid product: our free robots.txt generator plus a quarterly review covers it. Where paid enforcement earns its keep: bots that ignore robots.txt entirely.

Job 4 — Traffic measurement: "who is actually here, and is it working?"

The ground-truth layer: which AI crawlers, search indexers, and assistants actually visit, which pages they read, whether claimed identities are real or spoofed, and whether AI channels send you referral visitors your analytics buries. This is the layer we build — VisitorType classifies every visit (human / AI crawler / AI search / assistant / agentic browser), verifies claims against vendor IP ranges, and lets you fire different tags per type. What we deliberately don't do: answer tracking (job 1) or content scoring (job 2). One tool doing four jobs shallowly loses to two tools doing their jobs well.

What you actually need, by situation

• Just starting with GEO: job 4 first — measurement before optimization, or you're optimizing blind. Add the manual answer-tracking loop. Total cost: one tool + a spreadsheet.
• Content-heavy site, AI visibility is the strategy: jobs 4 + 1, with job 2 as an editorial checklist.
• Licensing/legal concerns about training data: job 3 enforced at the CDN, informed by job 4's data on who actually ignores your robots.txt.
• Agency running GEO for clients: all four, but insist on tools that export — the deliverable is the narrative across them, not four logins.

The category will consolidate; today's honest advice is to buy jobs, not suites. And whichever tools you pick, apply the same test we'd want applied to us: does it show its evidence (verified how? measured where?), or does it show you a proprietary score and ask for faith? In a field this young, evidence is the only moat that's real.