How to Verify an SEO Agency's Case Studies and Claims

Every agency shows you a case study. Almost nobody checks one. That asymmetry is why the same three moves keep working on smart, skeptical buyers: a percentage with no baseline, a 'rebuild' of pages that already existed, and a glowing reference whose day-to-day contact quit the agency two years ago.

You do not need SEO expertise to catch any of it. You need 15 minutes, a browser, and one question the salesperson did not rehearse. Here is the method. Run it on every deck on your desk — including ours.

How do you check whether an agency's case study is real?

Four checks, roughly 15 minutes, all free: confirm the absolute baseline behind the percentage, check the client's archived site against the build claim, check the traffic curve against a third-party estimate, and check that the reference's contact still works at the agency. A case study that survives all four is probably real. Most do not survive the first.

Notice what is not on that list: asking the agency for more detail. They will happily give you more detail. Detail is cheap and unfalsifiable. Independent corroboration is neither.

What does '+340% traffic' hide when the baseline was 11 visits?

It hides the only number that matters. Growing 11 monthly organic sessions to roughly 48 is a genuine +340% — and it is also a rounding error against any real service business's pipeline. Percentages are how a tiny absolute number gets dressed up as a result.

So ask one question, in writing: what were the absolute numbers, before and after? Sessions, phone calls, booked jobs. Not percentages. And not 'impressions', which climb happily when Google starts showing you for junk queries you will never rank for and nobody wants.

This is not just a marketing nitpick. The FTC's Endorsement Guides state that endorsements claiming specific results are usually read to mean the endorser's experience reflects what others can expect — and that saying 'Results not typical' or 'Individual results may vary' will not change that reading. The advertiser must either have proof the results are typical, or clearly disclose the generally expected performance.

The second trick is the date range. An HVAC case study measured from January to July captures the cooling-season ramp that would have happened with no SEO at all. Ask for year over year — July 2025 against July 2024 — not trough-to-peak. Then ask what a lead is worth, which is the whole point of measuring SEO ROI properly.

How do you use the Wayback Machine to test a build claim?

The Wayback Machine holds more than 1 trillion archived web pages, and it will show you what the client's website looked like the month before the agency started. It is the cheapest lie-detector in marketing, and it takes five minutes.

• Paste the client's domain into web.archive.org and open the calendar view.
• Jump to a snapshot from the month before the engagement started — the deck usually gives you the start date.
• Look for the specific pages the agency says it built: service pages, city pages, the blog, the new homepage.
• Compare the template. A real rebuild changes the navigation, the layout, and the URL pattern. A content refresh does not.
• Check when the 'new' site actually launched. If it went live two months before the engagement, that was somebody else's work.

Two honest limits. Coverage of small local sites is patchy, so a missing snapshot proves nothing — absence of evidence is not evidence of fabrication. And archived pages often render without their stylesheets, so 'ugly' does not mean 'old'. Use the archive to check existence and dates, not design quality.

What it catches reliably: a claimed redesign on a site whose template never changed; 'we built 40 city pages' when 38 of them sit in a snapshot from three years earlier; a portfolio site the agency never touched.

How do you sanity-check a traffic claim in five minutes?

Run the client's domain through a free estimator — Ahrefs' Website Traffic Checker returns a search-traffic estimate for any site — and look at the shape of the curve across the claimed window, not the absolute number.

Third-party estimates are estimates. They miss real traffic and invent traffic that is not there, so never use one to price a result or to argue about a specific month. Use it for exactly one job: does a spike exist where the case study says a spike exists? A flat line under a hockey-stick claim is a question worth asking out loud.

The only source of truth is Google Search Console. Its Performance report shows clicks, impressions, CTR and average position straight from Google, and it has a Share button — Google's own documentation notes the link grants access only to the current view of the report. So ask for a live screen share of the client's property, full date range, no filters, rather than a screenshot. Screenshots are trivially edited; a live GSC window is not.

If the agency says the client will not allow that, fine — then ask them to walk you through their own reporting on a call, unfiltered, with the date picker in your hands. An agency that will not let you touch the date picker is telling you where the cherry-picking is. A proper SEO audit starts from the same raw data, not from a slide.

What four reference-call questions cannot be coached?

Four questions, and the reason they work is that the answers live in the client's memory rather than the agency's script. A prepped reference can recite results. They cannot fake the texture of a working relationship.

Then ask the agency, not the reference, for one more thing: a client who left. Every reference list is a survivor list. An agency that cannot hand you a single churned client either has none — impossible after a few years — or is deciding for you which reality you get to see.

This matters most at the bottom of the market, where the deck is the entire product. If you are weighing a $500-a-month retainer, read what that money actually buys before you read another case study.

What does an anonymized case study legitimately look like?

Anonymity is normal and not a red flag — a personal injury firm in a competitive metro will not let a vendor name it. What is a red flag is anonymity used to strip out every checkable fact. A legitimate anonymized case study still gives you five things:

• The vertical and the metro size — 'a five-attorney firm in a top-20 US metro', not 'a client'.
• The exact date range, month to month, and a year-over-year comparison for anything seasonal.
• Absolute before-and-after numbers: sessions, calls, booked consults, and the spend that produced them.
• What the money bought — the channel mix, the number of pages, the link work, the ad budget.
• What did not work. A case study with no failed experiment in it was written by marketing, not by the person who did the work.

And it comes with an offer: a reference call under NDA. If the 'case study' is a stock photo, a logo you cannot verify, and one green arrow, treat it as advertising, not evidence.

Fabricated proof is also a legal problem now. The FTC's Rule on the Use of Consumer Reviews and Testimonials went into effect on October 21, 2024 and authorizes courts to impose civil penalties for knowing violations. The invented testimonial on an agency's homepage is not a grey area.

Point all of this at us too. We publish our pricing, we work month to month with no minimum, we do not guarantee rankings — anyone who does is lying — and you own the accounts, the content and the code from day one. That is the posture we would demand if we were the ones writing the check. It also happens to be how you should read a law firm SEO company's pitch, or any other vertical specialist's.

What should you do when the agency will not name a single client?

You have two options: walk, or make the case study irrelevant by changing the shape of the deal. If you cannot verify past results, stop buying on past results. Buy on a structure that lets you leave cheaply the moment results do not show up.

• Month to month, no minimum term. A 12-month contract protects the agency, not you.
• You own the ad accounts, the content, the links and the codebase from day one — not at the end of the term.
• A written 90-day checkpoint with a defined kill switch: no qualified leads, the channel gets cut.
• Reporting in booked calls and closed revenue. Impressions and rankings are diagnostics, not results.
• No ranking guarantee, in writing. Nobody controls Google's index.
• Published pricing. 'Call for a quote' means the number depends on how you look.

A twelve-month lock-in attached to an unverifiable case study is the most expensive combination in this industry. Get either verifiable proof or a cheap exit. Getting neither is how the last agency happened to you.

Run the four checks on every deck you have been handed, then run them on ours. Get my free audit and you will get exactly the specificity this post told you to demand: named pages, absolute numbers, and a plan you could hand to a competing vendor tomorrow.