TL;DR: AI SEO tools fall into four categories — crawlers, visibility trackers, content optimizers, schema generators — and they answer four different questions. Buy the crawler first, because a visibility score of zero cannot distinguish "the AI never saw your page" from "the AI saw it and chose someone else." Buy one tracker second, configured deliberately. Everything else can wait.
I am James, CEO of Mercury Technology Solutions. These products get marketed as AI SEO software, AI visibility platforms, GEO tools, LLM trackers — the labels move faster than the capabilities. Organize by the job each tool actually does, not by what the category is called this quarter. Here's the buying order we'd use if the budget were ours.
The Four Categories — Four Different Jobs
Category
What it does
What it tells you
What it cannot do
Crawler
Checks whether pages can be reached and parsed
Where access, indexing, and technical barriers sit
Decide what your brand should claim; guarantee AI visibility
Visibility tracker
Runs a defined prompt set, records appearances and citations
How you showed up for that prompt set
Explain why a result changed; represent every real user query
Content optimizer
Compares your content against topics, entities, competitors
Where coverage may be thin or duplicated
Judge whether information is useful, accurate, or worth including
Schema generator
Produces structured data markup from inputs
Whether valid markup can be generated
Verify that an entity relationship is true or matches the page
A crawler asks: can the system reach this page? A tracker asks: what happened when we ran these prompts? An optimizer asks: what could we add? A schema generator asks: can this be expressed in machine-readable form? Four useful questions. Buying one doesn't answer the other three — which is why the buying order follows the same dependency sequence as a GEO audit: crawler first, tracker second.
Question 1: Does the Vendor Start With Access, or With a Score?
A tracker is tempting because its output is legible: prompts in, score out. But a zero is not a diagnosis. If your content is missing from AI answers, two very different explanations exist — the content was accessible but not selected, or the crawler never reached it. The first calls for content and entity work. The second calls for robots.txt, rendering, status codes, and indexing fixes. Same zero, opposite responses.
Google's technical requirements are deliberately basic: Googlebot unblocked, successful HTTP response, indexable content — and even then, indexing isn't guaranteed. Passing technical checks gets you in line; it doesn't guarantee a seat. Technical SEO crawling comes before interpreting any visibility dashboard. It's why our own GEO Audit starts at the access layer, and why we publish how the audit is run instead of asking you to trust a black box.
Question 2: Which "AI Visibility Setting" Do They Mean?
Three levers exist, and vendors routinely blur them:
- Google-Extended (robots.txt) — controls AI training and grounding use of your content. Google states plainly it doesn't affect Search inclusion and isn't a ranking signal.
- nosnippet / data-nosnippet / max-snippet / noindex — how much of your page can be shown in Search and its AI features.
- Search generative AI control (Search Console) — whether you appear in Search's generative AI features at all.
If a vendor offers to "manage your AI visibility settings," ask which of the three they mean. If they can't name the right one, don't hand them your access layer.
Question 3: What Is the Tracker Actually Sampling?
"AI visibility" sounds broader than it is. A tracker only measures the prompts it was asked to run. What matters isn't whether the score is accurate — it's how it was produced: who chose the prompts, how many runs each, which language, region, and engine. A single run is an observation, not a finding. And scores across languages aren't directly comparable — an English prompt and a Traditional Chinese prompt retrieve different documents, surface different entities, cite different sources. That's not a translation issue; it's a different retrieval pool. Most AI SEO tools sample from US infrastructure in English by default — prompts about a Hong Kong purchase, sent from the US in English, measure a market you don't sell in. Configure sampling region and query language deliberately, and report them separately.
Question 4: Does the Optimizer Confuse Overlap With Usefulness?
Content optimizers have a legitimate use: coverage analysis across a large library surfaces genuine gaps fast. The problem starts when coverage becomes the goal. These platforms lean on semantic overlap because usefulness is hard to quantify and overlap is easy. But a page can mention every related concept and still be weak.
"Many businesses are investing more in AI search" and "a B2B SaaS company cut its publishing cycle from six weeks to three after restructuring approvals" overlap topically. Their commercial and editorial value are not equivalent — the second is verifiable evidence; the first is weather. A coverage score can't tell you which. This is where GEO programs waste budget: optimizing overlap instead of evidence. The better question isn't "how do we raise the score?" It's "what would make this page more useful, more defensible, more worth citing?"
Question 5: Who Decides What the Schema Says?
Schema generation looks finished when the validator passes. Valid syntax is the baseline, not the strategy. A generator can express an author relationship; it cannot decide whether the named person actually wrote the article. It can generate product markup; it cannot verify the markup still matches the page. Markup that contradicts visible content is how brands accumulate Truth Debt — machine-readable claims that erode trust instead of building it. The clean workflow: decide the entity → verify the relationship → make it visible → mark it up → validate. Not: generate everything → validate syntax → publish.
Question 6: What Stays on Your Desk No Matter What You Buy?
Three jobs no license covers: deciding what your brand should claim, producing evidence worth citing, and judging whether the numbers moved beyond the noise floor. Tools sample, compare, and format. Strategy, evidence, and judgment stay with you — or with a partner whose GEO methodology you can read. If you want the measurement discipline before the tooling conversation, the LLM KPI framework is the protocol layer, and the free GEO Scorecard shows you what a baseline looks like before you spend a dollar on software. That's the difference between buying an AI SEO tool and running an AI SEO program.
Stop buying dashboards. Start buying answers to the right questions — in the right order.
Mercury Technology Solutions: Accelerate Digitality.
