Most agencies are good at their work and honest about it. These six questions are not traps — they are how you find the one whose answers you can check, and a good agency will respect you for asking.
Hiring an SEO agency means buying expertise you cannot yet judge. The results arrive months after the first invoice, and the numbers that describe those results usually arrive in the agency's own reports. None of that is a reason for suspicion; it is a reason for clear questions at the start, while both sides are still deciding whether to work together.
The questions below all work the same way. Each one has an easy answer when the work behind it is sound, so they cost a good agency nothing — and each one sets an expectation that protects both sides. You know exactly what you are buying, and the agency knows exactly what it will be measured on, which is a better position for them than a client with vague hopes. If you are still deciding whether to hire at all, what SEO costs covers the money side; this piece is about choosing well once you have decided.
Where do your ranking numbers come from — which location and device?
Rankings are not one number: the same search returns different results depending on where the searcher is and what device they use. A rank tracker configured for the wrong location returns plausible-looking figures from a market you do not sell in, and nothing in the report reveals it. A good answer names the exact location and device settings behind every ranking figure, and shows them to you without hesitation.
This is not a hypothetical failure. We have written before about how a rank tracker set to the wrong location quietly measures a different country — the numbers arrive on schedule and look entirely normal; they are simply about searchers who will never be your customers. It is a configuration mistake, not a character flaw, and it can happen to anyone who sets up a tracker in a hurry. That is precisely why the question is worth a minute in the first meeting: once the settings are named, every ranking figure you see afterwards has a known address.
Which numbers in your reports are measured, and which are estimated?
Third-party tools estimate traffic and search volume from models, and for small sites those estimates can be off by multiples in either direction. Your own Search Console data is measured: those clicks actually happened. A good answer distinguishes the two clearly, labels them in every report, and never blends an estimate and a measurement into one column.
The estimates are not dishonest. Modelling is the only way any tool can describe sites it has no inside access to, and for large sites the models are often reasonable. The trouble is at the small end, where a modelled traffic figure can be several times too high or too low without the tool being broken. Search Console is different in kind, not merely in quality: it is Google's own record of your impressions and clicks — a log of what happened, not an estimate of it. A report that keeps those two kinds of number visibly apart is telling you something reassuring about everything else in it.
How do you report what you couldn't measure?
Honest measurement produces unknowns: a result that could not be read that day, a keyword the volume data source returns nothing for, a market the tool does not cover. A report with no unknowns anywhere has usually replaced them with something — a guess, a zero, or silence — and that is a quiet warning sign. A good answer describes how gaps are shown, not how they are avoided.
This sounds like the softest question of the six and is quietly the most revealing about method. Anyone who measures real things runs into things they cannot see, and there are only two ways to handle that: say so, or paper over it. An agency that writes "we could not measure this, and here is why" in a report is showing you the same discipline it applies to the numbers it can stand behind. Note that an unknown is not a failure — a missing number honestly reported is worth more than a confident number nobody can trace.
Which searches will you target — and would a buyer type them?
Traffic can grow month after month while every new visitor is researching rather than buying, and the report will look like winning the whole time. A good answer talks about buying-intent searches — the queries typed by someone ready to purchase — and asks what you actually sell before promising any volume.
The failure mode here is comfortable, which is what makes it persistent: a sessions chart that climbs for a year while enquiries stay flat, because the new visitors are students, researchers and competitors. We have described this pattern in traffic that looks like winning. The protection is decided before a single page is written, in the choice of searches to target. An agency that asks detailed questions about what you sell, who buys it and what those people would type is building towards customers. Volume can be added to a strategy later; intent cannot.
What will you not do?
Some practices that produce quick movement — bought link schemes, doorway pages, cloaking — violate Google's spam policies, and any penalty they trigger lands on the site owner, often after the contract has ended. A good answer names its own limits unprompted, because those limits protect both sides.
This question is about risk that outlives the engagement. The practices in question can move rankings for a while before they are detected, and when the consequence arrives it attaches to the site, not to whoever did the work. Google publishes its spam policies openly, so there is no ambiguity about where the lines are — only about whether a given supplier stays inside them. The encouraging fact is that most do, and the ones who do tend to be pleased to say so, because a stated limit is also a promise about the durability of the work you are paying for.
Do we keep our own Search Console access?
The data about your site belongs to you. Your business should own the Search Console property and grant the agency access to it — not the other way round — so that when the engagement ends, the history, the verified property and the measured numbers stay with you. A good answer treats this arrangement as obvious.
In practice this means the property is verified under an account your business controls, and the agency is added as a user with the access it needs. The direction matters at exactly one moment: the day the engagement ends, whether happily or not. Set up the right way round, that day costs you nothing — you remove one user and keep every year of history. Set up the wrong way round, your own performance record leaves with the contract. Agencies that work transparently prefer the right arrangement too, because it means their results remain visible to you in your own data long after the invoice is paid.
What success looks like
Success is enquiries and customers, not sessions. Sessions are a means; the end is someone who wants what you sell finding you and getting in touch, and the honest way to talk about that is in money — we walk through the arithmetic in what SEO is worth in money. An agency that frames its own success in your enquiries has aligned itself with you before the work begins.
Expect timelines in months, not weeks, and expect them without guarantees. "Guaranteed #1 rankings" is a promise nobody controls, because the results are Google's to give, not the agency's — so an agency that declines to guarantee a position is being accurate, not evasive. The realistic version sounds less exciting and is worth far more: which searches will be targeted, what will be built or fixed, when the first measured movement is plausible, and how you will both know.
Take the six questions to the first meeting, in writing if that feels easier. You are not setting a quiz; you are agreeing on how the work will be measured, which is the strongest start a working relationship can have — and the agencies worth hiring answer all six without breaking stride. TrustCtrl shows your own measured Search Console numbers alongside its site checks and labels every estimated figure as an estimate, so whichever agency you choose, you hold your own copy of the ruler.