We pointed our own SEO tool at our own site. It found five mistakes.

Most agency blog posts about SEO are written backwards. Something worked, so the story gets told as though the plan was obvious. You never see the part where the agency was wrong for six weeks.

This isn't that. We build an SEO plugin, we ran it against our own website, and it caught us making the exact mistakes we sell a tool to prevent. We're publishing it now, before the fixes have had time to work, because a method you only hear about after it succeeds isn't a method — it's a highlight reel.

First, the number that reframes everything else

Our domain authority is 0 out of 1000. The domain is fifteen years old. It has had several versions of a website on it. It has 39 referring domains.

We assumed that number was a bug. It wasn't. When we pulled the raw data, all 39 of those referring domains were link farms, URL shorteners, and directory spam — things like rankvance-agency.online and urls-shortener.eu. Every one of them scored zero. The oldest link in the profile dates to 2023.

The domain is fifteen years old. The links aren't. Domain age doesn't carry authority — links do, and ours are garbage.

That single fact reorganized our entire strategy, because it means content isn't our bottleneck. Authority is. No amount of writing fixes a competitive keyword when the sites ahead of you have thousands of real links and you have thirty-nine fake ones. Which makes the only interesting question: which keywords can be won without authority? That's what the rest of the audit was about.

Mistake 1: we optimized for competition and never checked demand

For about twenty-five pages, we chose keywords by asking one question: can we beat the sites currently ranking? We never asked whether anyone searched the term. Not because we decided demand didn't matter — the search volume column simply didn't exist in our tool, and nobody noticed it was missing.

When we added it, this is what came back:

We had called that first one the most winnable target on the whole site. Twenty-seven referring domains — practically undefended. It has zero searches a month.

And there's the lesson, which is obvious in hindsight and invisible in the moment: low competition and low search volume are usually the same fact. Nobody has built links for a phrase nobody types. We spent nine pages congratulating ourselves on an inverse correlation as though it were an edge. It was an empty room, and we mistook it for an open door.

Mistake 2: the phrasing gap, which turned out to be structural

This is the one we didn't expect, and it's now the most useful thing we know.

Pick any service keyword. Write down two or three ways of saying the same thing. The difference between those phrasings is routinely larger than the difference between entire industries:

Read those first two rows again. Identical demand — 880 searches a month, both. One is twenty-six times harder than the other. The restaurant pair is worse: roughly the same customer typing roughly the same thing, and about seven thousand times the competition.

This isn't luck, and it isn't a quirk of these two industries. It happened on every vertical we checked. Agencies crowd around the obvious phrasing — the one that sounds right in a pitch meeting — and leave the adjacent one almost empty. If you only ever check the phrasing that came to mind first, you will land on the crowded one nearly every time.

Mistake 3: a zero on one phrasing isn't a dead niche

We scanned four ways of saying "tow truck answering service." The first came back with zero searches, so we wrote off the whole niche and moved on.

Two of the other three were sitting in the same result set:

Better demand, a third of the competition, and advertisers paying fifty-nine dollars a click — which tells you the market is real even though the search box makes it look dead. We'd found it, scanned it, paid for the data, and skipped straight past it because the first number was a zero and we stopped reading.

Mistake 4: two of our pages were fighting each other

We had a page at /hvac-marketing/ and another at /hvac-lead-generation/. Same topic, same customer, competing with each other for the same searches — and the duplicate was aimed at the term with 8,257 rival links instead of the one with 317. We'd built a competitor for ourselves and pointed it at the unwinnable option.

Mistake 5: one word was worth 1,600 searches a month

Our Spanish marketing page targeted agencia de marketing digital: 1,300 searches, and out of reach for a site with our authority. The phrase agencia de marketing — the same thing, minus one word — gets 2,900 searches and is winnable. It was already in the page's headline, because it's a substring of the longer phrase. The fix was editing a single field.

That one stings, because the better keyword had been sitting in our own database for days. Already scanned. Already paid for. We only found it by auditing ourselves.

Mistake 6: we did it again, after we fixed the tool

This is the one that should have been impossible.

After adding search volume, after writing the rules, after auditing everything — we rebuilt our local SEO page and pointed it at local seo santa clarita. Thirty searches a month. We wrote “a real ranking play” into our own internal notes, with confidence.

Two hundred and seventy times the demand, sitting one phrasing away, still winnable — and we missed it the day after fixing the exact tool that exists to catch this.

That's the real lesson of this whole audit. Knowing the rule doesn't save you. We'd cited the phrasing rule five times that week and still walked past a 270× miss.

So we stopped relying on discipline and built a gate instead: the system now refuses to accept a keyword when a better one exists in the data. Not a reminder, not a checklist — a refusal. It would have blocked this one. Every mistake in this post was caught by data or by someone asking an awkward question. None of them were caught by us being careful.

What we changed

Eleven pages got demoted from ranking targets to conversion pages — they're aimed at terms nobody searches, so we stopped pretending otherwise. Several got retargeted onto phrasings we can actually win. The duplicate got merged. And the specialty terms we assumed were the prize — chiropractic marketing at $64 a click, dental implant marketing at $47 — turned out to be uniformly out of reach, because value and winnability are inversely correlated. The gold is guarded. That's not a market failure; that's the market working.

The part we can't tell you yet

Whether it worked.

Our site currently gets zero clicks from Google. Not "a few" — zero, across three months and 377 impressions. Every fix described above went live this month, and ranking takes months, so the honest status is: we don't know.

The one thing the data already told us is that our only visibility is local — every query we surface for at all has "Santa Clarita" in it. At zero authority, proximity is the one thing that outranks links. That's consistent with everything above, and it's why our next work is aimed there.

We'll update this post with what actually happens. Including if the answer is that it didn't work.