Your SEO Report Said Traffic Was Falling. The Traffic Was Fine.

We almost sent a client a report telling them four months of SEO work had gone backwards.

It would have been wrong. Not slightly wrong. Wrong in the sense that the decline we were about to explain did not happen.

Here is the whole thing, including the part where we were confidently incorrect, because the lesson is more useful than the diagnosis.

What the report said

A senior living operator we work with, communities across two states. We are a few months into a technical SEO push.

Google Search Console showed impressions sliding. Top-three rankings had dropped. The month just closed looked like the worst in a while.

So we went looking for a cause, and we found one fast.

There were two versions of the same page in the data. One singular, one plural. Same content, same purpose. Both showing impressions, going back months.

That is a textbook problem. Two URLs competing for the same query, splitting authority, neither one winning. Keyword cannibalization.

It got better. The old singular URL was frequently sitting in position one or two. The live plural page was down at four, six, eight, twelve depending on the community.

So the story wrote itself. Google was still serving a dead page. The redirect was not being honored. Directory citations were feeding the old URL. Roughly 1,800 monthly clicks were being split across duplicate pairs.

Clean narrative. Clear villain. Obvious fix.

All of it was wrong.

The cross-check

Before writing any of that into a client report, we checked it against other sources. Not because we doubted it. Because that is the rule.

Four sources, independently:

• Ahrefs. The singular URL was not ranking anywhere.
• Live SERP data. We pulled the actual search results. The plural URL was there. The singular URL did not appear in the top 20 for anything.
• Google Analytics 4. The old URL had zero pageviews. Not a handful. Zero, across twelve months.
• The URL itself. We inspected it. Proper 301, redirecting correctly, and it had been doing so since 2021.

Every source disagreed with Search Console.

The redirect had never failed. Google was not serving a dead page. There was no cannibalization. Nothing was splitting 1,800 clicks, because the old URL was not receiving clicks at all.

Search Console was reporting against a legacy canonical record. It was showing us its internal bookkeeping, not what Google actually puts in front of a family searching for care.

I had been doing this a long time and had not seen it behave that way before.

So what was actually happening

Once the ghost data came out of the picture, the real numbers were unremarkable.

Organic traffic was essentially flat. Comparing the two most recent months, the difference was about six visitors. Not a decline. Noise.

And some of the work was clearly landing. One community's assisted living page had moved from an average position around 37 to about 1.4. That is page four to the top of page one.

The four months of decline was mostly an artifact sitting on top of a roughly flat baseline, with real wins underneath it that the report had buried.

If we had sent the original version, we would have spent a client call apologizing for progress.

The part that stings

We had not reached that first wrong conclusion by hand. We had run the analysis with AI assistance, and it produced a confident, coherent, well-structured root cause.

It was articulate. It was specific. It cited the right-looking evidence. And it was built on a data source that was lying.

That is the failure mode worth understanding, because it is not going away. AI analysis inherits the quality of its inputs and then presents the result with total composure. It does not hedge. It does not say "this one source disagrees with the other three," unless you make it look at the other three.

Garbage in, confident garbage out, formatted beautifully.

We use these tools constantly and they have made us much faster. But speed applied to an unverified input just gets you to the wrong answer sooner, with better formatting.

The verification step is not optional overhead. It is the actual job.

The real root cause

There is a satisfying ending to this.

We eventually worked out why the ghost URL kept generating impressions years after the redirect went in.

Every one of the client's Google Business Profiles still listed the old URL in its website field.

We were feeding it ourselves. Dozens of authoritative listings, pointing at a URL that had not been the real address since 2021, quietly keeping a legacy record alive in Google's index.

We updated them on the call.

This is the genuinely useful, go-check-it-now part. Your Google Business Profile website field is one of the strongest signals you control. If it points somewhere stale, you are actively telling Google your old address is your real one, and you are muddying the entity picture that AI assistants use when a family asks who to trust.

If your site has been redesigned, migrated, or restructured in the last five years, go look. Every profile, every community, right now. It takes ten minutes.

The protocol

Here is what we do before anything goes in a report or a decision. It is the same discipline behind getting a business ready to use AI without getting burned by it.

• Never act on one source. Search Console, Analytics, a rank tracker, and a live SERP check. If they agree, the finding is real. If one disagrees with three, the outlier is usually the problem, not the finding.
• Check impressions against actual traffic. A URL with impressions and zero pageviews is not a ranking page. It is a bookkeeping entry.
• Verify the mechanism, not just the symptom. Before concluding a redirect is broken, inspect it. Before concluding a page is cannibalizing, look at the live results.
• Ask what timeframe you are comparing. Comparing a period contaminated by ghost impressions to a clean one produces a decline that never happened.
• Make AI show its disagreements. Ask what data would contradict the conclusion. Ask which sources were not checked. A tool that cannot tell you where its analysis is weak is not finished analyzing.

What to do with your own report

If your agency hands you an SEO report showing a drop, three questions.

Which independent sources confirm this, and do they agree?

Does the traffic data in Analytics match the impression data in Search Console?

What is the specific mechanism, and has anyone verified it rather than inferred it?

A good partner will have answers ready, and will occasionally tell you the scary number was an artifact. That is not spin. Knowing what your data actually says, including when it says nothing, is the part you are paying for.

The same instinct applies to the content you already have. Before you commission anything new, audit what you already published, because the wins are often sitting in pages you have already paid for.

We came very close to selling a client a crisis that did not exist. The only reason we did not is that we checked.

We work only with businesses that serve seniors: senior living communities, home care and home health agencies, senior placement advisors, elder law firms, and senior move managers. If you serve seniors, we serve you.

Want a second set of eyes on a report that does not sit right? Let's talk.