Tool, diagnosis, or neither?

Who is telling you this

The Citation Lab sells diagnostic work. Our reports run on tooling built by AuditSpark, and the guides in this section are adapted from AuditSpark's published research. That is a real interest in one of the three answers below, and you should read the rest of this page knowing it.

The reason to say so plainly is that a buying guide written by a vendor that never names its own stake is doing the thing this whole site exists to find in other people's marketing. So the test we would apply to someone else applies here: every recommendation below should still make sense to you if you assume we would prefer you pick ours. Two of the three sections say don't.

A score is not a diagnosis

The most expensive confusion in this category is treating a number as an explanation. A monitoring tool can tell you, accurately and continuously, how often engines name you. It cannot tell you which of the things that produce that number is the one holding you back, because that requires reading your site, your structure, and the off-site evidence about you, and then forming a judgment about which constraint binds first.

We can put real numbers on that gap, because they are ours. AuditSpark audited this site twice in August 2026, sixteen days apart. On 8 August it scored 93 out of 100 on AI and GEO readiness against 11 out of 100 on measured AI visibility, with an independent share of voice of zero. On 24 August, after a fortnight of work on the site itself, readiness had risen to 98 and visibility had not moved at all: still 11, still zero share of voice. Every on-site signal a tool can read is close to perfect. The thing the business actually wants, being named when a buyer asks, is close to absent.

That is not a broken measurement, and it is not a paradox. It is the split the visibility ecosystem describes: readiness is a property of your site, visibility is a property of engine behavior, and engine behavior leans heavily on evidence that lives somewhere other than your domain. A dashboard reporting both numbers would have shown us the 93, then the 98, and the 11 twice, and left the interesting question, which one can we move and how, entirely open.

This cuts against us as much as for us. If your constraint is off-site (nobody has written about you, no third party has corroborated you, you are absent from the places engines actually draw on), then a diagnosis of your site will confirm your site is fine and hand you a problem that no amount of on-site work solves. That is worth knowing before you buy one.

When a subscription is the right buy

Buy monitoring when the thing you need is a time series, and when someone will look at it. Concretely, that means most of these are true of you:

• Your category moves. New entrants, active competitors, engines that have started naming someone else in the last quarter. If the answer to "who gets recommended in our space" would plausibly be different in ninety days, a snapshot ages badly and a trend line earns its cost.
• Somebody owns it by name. Not a team, a person, with the standing to change a page when the data says to. Tools do not act. The most common failure here is not a bad tool, it is a dashboard nobody has opened since the onboarding call.
• You have already done the structural work, or know that you have no structural problem. Monitoring is how you catch a regression and how you prove a change did something. It is a poor instrument for finding a defect you have never looked for.
• You want to watch competitors, not just yourself. Share of voice against a named set is genuinely easier to buy than to build, and it is the number most worth tracking over time.

If those describe you, buy the tool and do not hire anybody. Entry pricing in this category is published and easy to compare across vendors before you commit to any of them, which makes it a real cost but a legible one, and a cheaper one than the alternative if what you genuinely need is a number every week.

When a diagnosis is the right buy

Buy an assessment when the question is why, not what, and when you would not know what to do with a score. Typically:

• You have never checked, and you do not know if you have a problem. This is the most common position and the cheapest to resolve. A readiness check answers it in minutes at no cost, which is a good enough reason to do that before considering anything paid, including ours.
• You have a number and it is bad, and you do not know why. A score with no explanation is where a lot of tool subscriptions quietly stall. Working out which of the underlying layers is the binding constraint is analytical work, and it is the thing a dashboard structurally cannot do for you.
• Nobody in-house reads this stuff. If interpreting the output is going to fall to a person who has never thought about crawler policy or structured data, you are buying a tool and a research project. The research project is usually the more expensive half.
• The likely fixes are one-time and structural. Crawler access, schema, page shape, and trust architecture are largely things you fix once and maintain. You do not need a recurring subscription to tell you every week that the same robots.txt rule is still wrong.

The honest limit on this option is the one above: a diagnosis is a judgment at a moment. It ages, it cannot promise a citation, and if your real constraint is off-site then it will tell you so rather than fix it.

When to do nothing yet

This is the option nobody in the category writes a page about, and it is right more often than the other two combined.

• Your buyers do not ask engines about businesses like yours. The behavior is real but it is not uniform, and it concentrates in considered purchases where somebody is researching before they commit. If your customers arrive by referral, by walking past, or from a relationship built over years, the answer layer is not where your next customer is deciding.
• Something upstream is more broken. If your site does not explain what you sell, or takes eight seconds to load, or has no way to contact you, being cited by an engine sends a stranger to a page that was going to lose them anyway. Fix the thing that converts before buying visibility into it.
• You have no capacity to act this quarter. Information you cannot act on is not an asset. Buying either option to feel like you did something is the most expensive of the three, because it costs money and it also costs you the accurate belief that you have not addressed this yet.

Waiting is a defensible position with one condition attached: pick the date you will revisit it and write it down. The difference between deciding not to act and never having looked is entirely whether you set the review.

The check that costs nothing

Whichever of the three you are in, the sequence that wastes the least money is the same. Establish whether you have a problem before you buy an instrument for measuring it, and prefer the cheapest test that could change your mind. In practice: run a free readiness check to see whether engines can reach and read your site at all, then ask a handful of engines the questions your buyers ask and write down who gets named. That second step costs an afternoon and no money, and it is the one that tells you whether this is urgent, background, or genuinely not your problem.

If both come back clean, you have your answer and it is the third option. If the readiness check is clean and the engines still never name you, you are where this site is, and the work is off-site. If the readiness check is not clean, you have found a real defect and it is fixable. Only in the last two cases is there anything worth buying, and only in one of them is it us.

Frequently asked

Can a free checker tell me everything a paid assessment would?

No, and it is not meant to. A free readiness check reads what is publicly visible on your site: whether crawlers are permitted, whether your structure is machine-readable, whether there is answer-shaped content to lift. It cannot tell you what engines actually say about you, and it cannot weigh which of several real problems to fix first. It is a good enough instrument to tell you whether to keep going, which is all it should claim.

Do I still need monitoring after I have fixed my site?

Only if your category moves fast enough that the answer would change. Fixes to crawler access, structured data, and page structure are largely durable, so re-measuring weekly tells you very little. What justifies a subscription is competitive movement and engine behavior shifting under you, not the possibility that your own schema spontaneously breaks.

How do I know whether my buyers use AI to find businesses like mine?

Ask a small number of them how they found you and what they looked at before they got in touch, and put your own category's buying questions to two or three engines to see whether the answers name real competitors or nobody at all. If engines return a confident list of firms like yours, buyers can see that list. If they return something vague and generic, the behavior has not arrived in your category yet.