Independent search marketing consulting
Abstract vertical bar illustration representing Questions to Ask Before You Sign

EngagementDecisionA choice you have to make before you spend anything.

Questions to Ask Before You Sign

Short answer
Ask about the work, the person, the measurement and the exit — and judge the shape of the answer rather than its confidence
What you receive
A list you can take into any call, including one with me
Applies to
Consultants, agencies and in-house candidates alike

Questions with the good and bad answers written next to them, so the list works on anyone you point it at

How to use this, including on me

This list is written to be taken into a call with any search provider, and it is deliberately not written to make one kind of provider look good. If you use it on me and the answers are weak, that is information you should act on.

Two principles make it work. Judge specificity, not confidence. The most dangerous answers in this field are fluent, warm and unfalsifiable. A slightly awkward answer full of conditions is usually the honest one. And listen for what an answer would rule out. A statement that could not be wrong under any circumstances tells you nothing, however good it sounds.

You are not trying to catch anyone out. You are trying to establish four things: that they understand your problem, that you know who will actually do the work, that everyone agrees in advance how success will be measured, and that you can stop.

Questions about the work itself

What do you think is actually wrong, and how would you confirm it?

Good: a named hypothesis with a method of testing it — “your category pages look like an indexation problem; I would check coverage reports, whether the filters are generating crawlable URLs, and whether the canonical tags agree with your internal linking before saying anything more.” Bad: a list of general improvements that would apply to any website, delivered before anyone looked at yours.

What would you not do here, and why?

Good: a specific declination with a reason — “I would not touch crawl budget on a site this size; Google's own thresholds put that concern well above where you are.” Bad: a claim that every service offered adds value to every client. A provider with no refusals has no judgment, or is not sharing it.

Do you follow Google Search Essentials?

This is one of Google's own recommended questions and the answer should be immediate and slightly puzzled that you asked. Bad: any answer that treats the guidelines as optional, aspirational, or as something competitors get away with ignoring.

What results and timeframe do you expect?

Also Google's question. Good: ranges, conditions and dependencies, with different clocks for different work — technical fixes visible within weeks of recrawl, content and authority work over months, no assurance attached to either. Bad: a single confident number, or a date attached to a position.

Questions about who actually does the work

Who will do this work, and will they be on this call?

Good: a name, a role, and whether that person is present now. If the answer is “a senior strategist to be assigned”, ask when you will meet them and whether you can decline the assignment. Bad: a description of a team's collective experience with no individual attached, which is a common way of answering a different question than the one you asked.

What proportion of this is subcontracted?

Good: a plain answer. Subcontracting is not a problem in itself — content production and development frequently are, sensibly. Bad: discomfort, or a denial that later turns out to be technically true because the subcontractor is on a retainer rather than a contract.

What happens if that person leaves?

Good: a real answer about documentation and handover, and, from an independent, an honest statement about what happens to your engagement if they are unavailable. Bad: reassurance without mechanism.

How much of your time will this get each month, and doing what?

Good: an approximate split between analysis, specification, review and meetings, with the caveat that it varies by phase. Bad: a number of hours with no description of what fills them. Hours are an input; you are buying decisions.

Questions about measurement

What baseline will you record before starting, and when?

Good: a specific set — impressions, clicks and average position for the pages in scope on a stated date, with the query set defined, plus the conversion figures. Bad: a plan to start measuring later, or an assumption that the analytics you already have will serve.

What exactly counts as a conversion, and who defines it?

Good: the definition is yours, written down, identical across paid and organic, and matched to something in your own systems. Bad: the provider's platform defaults, or a definition that includes page views and time on site. If paid counts a lead at form submission and organic counts it at qualified opportunity, no comparison between them means anything.

What will you report, and what will you not be able to tell me?

Good: stated limits — Search Console retains 16 months and caps the interface at 1,000 rows; rare queries are suppressed so filtered totals never sum to unfiltered ones; average position means the topmost result, not an average of all of them; rank tracking is a synthetic sample rather than a universal position; and no source reports clicks from AI features. Bad: a single composite score. A proprietary “SEO score” reports the vendor's opinion of your site.

How will I know a recommendation actually shipped?

Good: a change log with dates, tied to releases or tickets. Bad: the assumption that recommended and implemented are the same event. They frequently are not, and this is the most common reason engagements fail invisibly.

Questions about links and anything being bought on your behalf

Where do links come from, and is money involved?

Good: a description of the actual mechanism — digital PR, published research, partnerships, reclaiming existing unlinked mentions — and, if anything is paid for, an explicit statement that those links carry rel="nofollow" or rel="sponsored". Google is clear that paid links qualified this way are not a violation; undisclosed ones are. Bad: a monthly link quota, a network of sites nobody will name, or an appeal to private relationships. A quota implies a supply, and a supply implies a purchase.

Will you show me every link placed, with the URL?

Good: yes, as a matter of course. Bad: a summary count, or a sample. You carry the risk of these links, so you should be able to see all of them.

Do you plan to use the disavow tool?

Good: probably not, with a reason — Google's own guidance states most sites will not need it. Bad: routine disavowing as a monthly service line, which sells work against a problem you likely do not have and can cause harm if applied carelessly.

Would I ever be asked to link to you?

Google's wording on this is unusually blunt: you should never have to link to an SEO. A provider requiring a footer credit link on your site as a term of service has an interest that is not yours.

Questions about the commercial terms

What is included, and what is expressly excluded?

Good: a written scope where deliverables are artifacts — a findings document, specifications, a change log, an account structure — and where content production and implementation are either included and priced, or excluded and named. Bad: deliverables written as activities. “Ongoing optimization” and “technical improvements” can be satisfied by nearly nothing.

What do you need from me, and what happens if I cannot supply it?

Good: a list of dependencies with owners — developer time, content approval, analytics access, a decision-maker who can settle disputes. Bad: nothing required from you, which means either the work will not touch your site or the dependency will surface later as a delay you get blamed for.

How does this end, and what do I keep?

Good: notice period stated, no penalty for stopping at a review point, and a plain statement that your analytics and Search Console properties, advertising account, content and documentation remain yours and accessible. Bad: automatic annual renewal, long notice periods, or an advertising account held under the provider's own management identifier with no transfer arrangement discussed.

What would make you end it from your side?

Good: an honest answer — persistent inability to ship, a client insisting on something against policy, a scope that keeps expanding. Bad: nothing. A provider with no exit conditions is describing a revenue line, not an engagement.

The answers that should end the conversation

Most weak answers deserve a follow-up question. Four do not.

  1. A guaranteed position, ranking or traffic figure, in any wording. Google states directly that no one can guarantee a number one ranking, and warns you off providers who claim it. This is not a negotiating position; it is a statement about how the system works.
  2. A claimed special relationship with Google, or a priority submission. Google names both explicitly as things to beware of. Neither exists.
  3. A refusal to explain what will be done. Google's guidance says to be careful if a company is secretive or will not clearly explain its intentions, and to insist on understanding every change made to your site. Trade secrecy in this field is almost always concealment of something you would object to.
  4. Any suggestion that a violation is safe because everyone does it. The consequence lands on your domain, not on the provider's.

One softer signal is worth as much as the four hard ones: if you ask a direct question twice and get a category-level answer both times, you have your answer. The people who do this work well are usually relieved to be asked something specific.

What to write down before you sign

Answers given warmly on a call have a way of not appearing in a contract. Five items are worth insisting on in writing, and none of them are unreasonable to ask of anyone.

  • The named person doing the work, and what happens if that changes.
  • Deliverables as artifacts, each with a date, plus explicit exclusions.
  • The baseline, recorded before work starts, with the metrics and the query set defined and the conversion definition agreed.
  • The change log commitment — that what shipped, and when, will be recorded, because everything you later want to know depends on it.
  • Ownership and exit — accounts, properties, content, documentation, notice period, and no penalty at a review point.

A provider who agrees to all five on a call and to none in writing has told you which of the two conversations was real. Google's own hiring guidance covers the first half of this list and is a useful thing to have open while you read a proposal.

Frequently Asked Questions

What questions does Google recommend asking an SEO?

Google's hiring guidance lists six: can you provide examples of previous work and success stories; do you follow Google Search Essentials; what results and timeframe do you expect; what is your industry experience; how will you communicate changes and recommendations; and are you genuinely interested in understanding my business. The same page warns against guaranteed rankings, claims of a special relationship with Google, priority submission offers, cold outreach, link popularity schemes, and providers who will not explain what they intend to do.

What is the single most useful question to ask?

What would you decline to do here, and why. It cannot be answered well without judgment about your specific situation, it cannot be answered from a template, and the answer reveals whether the person is willing to talk you out of something. Experienced practitioners have ready examples — declining crawl budget work on a small site, declining content volume when the existing pages have not been fixed, declining an engagement where nobody can implement. A provider with no refusals is selling capacity, not advice.

Should I ask for references or case studies?

Ask, and then interrogate what you get rather than accepting it. Most published case numbers are bare percentages with no baseline, no date range and no stated measurement source, which makes them unverifiable by construction. More useful than a number is a described situation: what was wrong, how it was diagnosed, what was changed, what did not work. A reference conversation with a past client is worth more than any document, and the question to ask them is what went wrong and how it was handled.

How do I ask about pricing without getting a sales pitch?

Ask what drives the number rather than what the number is. A useful answer names structural drivers — whether implementation is included, whether content production is included, template count rather than page count, number of markets and languages, platform constraints, how many stakeholders must be brought along. Then ask what would make this cheaper, which forces a scope conversation. A provider who cannot explain the shape of their own pricing either has not thought about it or is pricing off what they think you can pay.

Is it rude to ask who is actually doing the work?

No, and the reaction to the question is itself informative. Agencies commonly staff a pitch with senior people and deliver with junior ones, which is a normal business model rather than a scandal, but you should know which you are buying. Ask for a name, whether that person is on the call, what proportion of the work is subcontracted, and what happens if they leave. Providers who do this well answer immediately because they have been asked many times.

What if the answers are good but something still feels wrong?

Buy something small first. A bounded diagnostic project produces an artifact you keep, tests how the person works, reveals whether they write specifications your team can actually build, and costs a fraction of a year's commitment. If it goes well you have a working relationship and a baseline. If it does not, you have learned something cheaply, and you still own the document. Uncertainty is a legitimate reason to reduce scope rather than to delay indefinitely.
Keep reading

Read the guides

An entry states what a rule requires or what a dispute turns on. A guide walks the sequence — what you do, in what order, before the evidence is gone.

Top