Answered straight

How to choose an SEO agency without getting burned

The short answer

Choose an SEO agency by testing it, not by reading its deck. In one 30-minute call, ask for a campaign that failed, the name of the person who'll do the work, how they'd define your baseline, where their links come from, and what they refuse to promise. Then call a client who left.

Updated 26 July 2026 · Written by the Last Agency team · See what SEO actually costs

The short version

  • A pitch deck is optimised to pass. Design a test it can fail, or you're just comparing slide quality.
  • The two most informative questions in the whole call: *tell me about a campaign that didn't work*, and *who exactly does the work?*
  • Every test below is answerable in under two minutes by an honest agency. The hedging is the data.
  • Two hard fails in one call is a walk-away, however good the chemistry was.

The pitch is built to pass. Build a test it can fail.

Every agency you speak to has run this call a hundred times. The deck is polished, the case studies are the three best ones they've ever had, and the questions you were planning to ask have scripted answers waiting.

So stop evaluating the pitch and start running tests. A test is a question where the wrong answer is obvious to you and impossible to fake in real time. Nine of them fit comfortably into a 30-minute call, and you don't need to know any SEO to score them.

Take notes as you go. What you're listening for isn't expertise — it's whether the answer arrives quickly and specifically, or slowly and generally. An agency that has thought about a question answers it in ten seconds. One that hasn't takes two minutes and never quite lands.

Tests 1–3: the honesty tests

These are first for a reason. If an agency fails here, the rest of the call is theatre.

Test 1 — "Tell me about a campaign that didn't work."

Nobody has only won. Every agency has an account that went sideways, and the ones worth hiring can tell you about it without flinching, because they learned something specific from it.

  • Pass: a real story with a named cause — a client who never shipped the technical fixes, a market where the incumbents' link profile was unbeatable at the budget, a migration that went wrong on their watch. Bonus points for taking the blame.
  • Fail: "we've never had a client leave unhappy", or a failure story that's secretly the client's fault in every detail.
  • What it reveals: whether they diagnose their own work. An agency that can't name a failure can't name a cause, and will do exactly the same thing to you.

Test 2 — "What will you refuse to promise me?"

This one is deliberately awkward, and the discomfort is the point. It asks a salesperson to volunteer a limit mid-pitch.

  • Pass: a fast, clear list. No specific ranking positions. No timeline shorter than three months for competitive terms. No control over what Google ships next quarter.
  • Fail: "we can pretty much get you anywhere", or a promise of a #1 spot. Google's own guidelines flag ranking guarantees, and no agency controls the index — see whether SEO agencies can guarantee rankings.
  • What it reveals: whether they've ever had to explain a limit to a client. The ones who have will do it in ten seconds.

Test 3 — "Give me a reference from a client who left."

Anyone can produce three delighted current clients. A churned client is the reference that carries information: why they left, whether it ended cleanly, whether the handover was honest.

  • Pass: they name one, or explain candidly why they can't — a signed NDA, an acrimonious exit they'd rather describe themselves than send you into.
  • Fail: visible panic. Or a claim that no client has ever left, which is either false or means they've been trading for eleven months.
  • What it reveals: how they behave when the money stops. That's the only version of an agency's character that matters to you six months in.

Tests 4–6: the delivery tests

Now find out what actually gets built, and by whom. This is where the gap between agencies is widest and the pitch is least reliable.

Test 4 — "Who does the work, by name, and will they be on my calls?"

The people who win the pitch are very often not the people who do the work. It's the single most common complaint founders bring us about their previous agency, and it's entirely preventable at the contract stage.

  • Pass: actual names, actual roles, and a straight answer on whether the work is done in-house or subcontracted. Subcontracting isn't automatically disqualifying — hiding it is.
  • Fail: "you'll have a dedicated account manager". That describes the person who forwards emails, not the person who writes or fixes anything.
  • Follow-up worth asking: how many accounts does that named person carry? Twelve is a very different answer from four.

Link building is where agencies most often buy a shortcut with your domain as collateral. The request is simple: live URLs from a recent month, for a real client, in whatever redacted form they need.

  • Pass: named publications, digital PR angles, expert commentary, partnership placements. They can describe the pitch that earned each one.
  • Fail: "our network", "our relationships with high-DA sites", or a rate card by domain authority tier. A price list per DA band is a paid-link business with a spreadsheet.
  • Why it matters: paid links violate Google's spam policies, and the consequences land on your site, not their invoice.

Test 6 — "What ships in month one?"

Month one is where vagueness hides. A good agency has a near-identical answer for every client, because the first month is genuinely the same shape every time: audit, baseline, plan, and the first technical fixes.

  • Pass: a specific sequence with deliverables and dates. Crawl and audit in weeks one to two, baseline frozen and signed off, technical fix list prioritised, content plan by week four.
  • Fail: content going live in week one. That means they're writing before they've researched, which is how you get 40 articles nobody searches for.
  • Also fail: an answer that's mostly onboarding admin. Access setup is a task, not a deliverable.

Tests 7–9: the money tests

Three questions about what you're buying and what happens if it doesn't arrive. These decide whether the relationship survives month four.

Test 7 — "What's my baseline, and how will you measure it?"

If nobody writes down where you started, nobody can prove you moved — and every monthly report becomes an argument about which chart to look at.

  • Pass: a business number, frozen before work starts. Qualified leads from organic search over the trailing 90 days, or organic revenue. Recorded, agreed, dated.
  • Fail: "we'll track rankings and traffic." Both are trivially inflatable — rank for enough irrelevant long-tail and the charts go up while your pipeline doesn't.
  • Practical step: set the baseline yourself before the first call, so you can compare what each agency proposes to measure.

Test 8 — "How long is the contract, and how do I leave?"

Contract length is the clearest statement of an agency's confidence. A 12-month lock-in with a 90-day notice period says the revenue matters more than the result.

  • Pass: a short initial commitment — a quarter is fair, since nothing meaningful can be judged sooner — then month-to-month with 30 days' notice.
  • Fail: any lock-in longer than six months, auto-renewal clauses, or an exit that requires a quarter's notice.
  • Ask separately: what do I keep on exit? Accounts, content, data and access should all be yours. Get it in the contract, not the email. What belongs in an SEO contract covers the rest.

Test 9 — "What happens if it doesn't work?"

The question that separates the field, and the one most agencies have never been asked. Watch for whether the answer contains a consequence for them, or only for you.

  • Pass: a defined trigger, a defined response. Continued work at no charge, a scope change at no charge, or a clean exit. Tied to the baseline from Test 7.
  • Fail: "SEO takes time." True, and also the answer given by every agency that has never carried any risk.
  • Ours, for comparison: we freeze your trailing-90-day qualified organic leads on day one, and if we haven't beaten it in 90 days we keep working free until we do. That's a guarantee of movement against your own number — never a promise of a specific ranking position, which nobody can honestly make.

Scoring the call

Score each test pass, soft or fail immediately after the call, before the chemistry wears off. Chemistry is the most persuasive and least predictive input in the whole process.

The weighting isn't equal. Some fails are survivable; two of them end the conversation regardless of everything else.

How to weight each of the nine tests when comparing agencies.
TestWeightIf they fail it
1. A campaign that failedHighThey don't diagnose their own work. Serious concern.
2. What they refuse to promiseDisqualifyingA promised ranking position ends the call.
3. Reference from a churned clientMediumSurvivable with a credible reason. Panic isn't a credible reason.
4. Who does the work, by nameHighYou're buying a salesperson, not a team.
5. Where links come fromDisqualifying"Our network" is a penalty with a payment plan.
6. What ships in month oneMediumFixable — ask for the plan in writing before signing.
7. Baseline definitionDisqualifyingNo baseline means no way to prove anything, ever.
8. Contract and exitHighNegotiable, and their willingness to negotiate is the real answer.
9. What happens if it doesn't workHighYou're carrying all the risk. Price accordingly.

Three things that look like evidence and aren't

Most shortlists are built on signals that carry almost no information. Worth knowing which ones to ignore before you spend a weekend comparing them.

  • Awards and directory rankings. Many agency directories sell sponsored placement and rank partly on review volume, which agencies solicit. Read why 'best agency' lists are paid before treating a badge as third-party validation.
  • Client logos. A logo wall tells you someone once signed an invoice. It doesn't say what the scope was, whether it worked, or whether the client is still there. Ask which of those logos is a current client and watch the list shrink.
  • Tool stacks and certifications. Everyone has the same three tools. A Google certification is a free online exam. Neither predicts whether anyone at the agency can write a page that ranks.
  • A free audit. Usually a tool export with a sales call attached, built to find alarming-sounding problems rather than the ones costing you money. Useful as a writing sample, not as a diagnosis.

Related questions.

What should I ask an SEO agency before signing?

Nine things, all answerable in one call: a campaign that failed, what they refuse to promise, a reference from a client who left, who does the work by name, where their links come from, what ships in month one, how they define your baseline, contract and exit terms, and what happens if it doesn't work.

How much should I expect to pay an SEO agency in India?

Serious multi-discipline work generally starts around ₹75,000 a month, with smaller or local sites viable from ₹40,000. Below ₹25,000 you're funding part-time attention rather than a team. Our own SEO starts at ₹75,000/mo, and ₹40,000/mo for smaller sites, all ex-GST.

Should I choose a specialist or a generalist agency?

Specialism in your channel matters far more than specialism in your industry. An agency that has done ecommerce SEO can learn your category in a fortnight. One that mostly runs ads will learn SEO on your budget. Ask which discipline they'd lose sleep over, then check that's the one you're buying.

How long should an SEO contract be?

A first quarter is fair, because nothing meaningful can be judged in under 90 days. After that, month-to-month with 30 days' notice. Any lock-in beyond six months protects the agency's revenue rather than your outcome, and auto-renewal clauses are worth striking out entirely.

Is a free SEO audit a useful way to choose an agency?

Only as a writing sample. Free audits are sales documents — usually a crawler export with the scariest findings pushed to the top. Judge it on whether the recommendations are prioritised by likely business impact or just listed by severity score. Most are the latter.

How many agencies should I speak to?

Three. One tells you nothing because you have no comparison. Five turns into a procurement exercise where the best deck wins. Three, scored on identical questions within the same fortnight, gives you real contrast while the answers are still fresh enough to compare.

Last slot's open

Make this the last growth call you book.

Grab the free strategy call and walk away with a 90-day growth plan — hired or not. Or just text us. Either way, you'll know exactly how we'd win.

Guaranteed or it's free · No lock-in · Free strategy call