Transcript
Intro
A patent and trademark firm, in business since 1999, got a shiny new website. Another team moved it from ASP.NET to WordPress, and nobody set up redirects. Every old URL turned into a 404, and organic traffic slid for a long time. No penalty. No Google update. Just a missing list of redirects.
Then it came to us. We rebuilt it again and redirected every old URL from both eras. In under three months, organic traffic went from about 400 visitors a month to nearly 4,000. Plus 830 percent. We'll show you how at the end. First, how to work out what actually went wrong when your traffic falls.
Here's the order. Six questions, asked in sequence, about thirty minutes, and they stop you fixing the wrong thing. Four of them are on screen now.
Then two more: what did we change ourselves, and only at the very end, is it a quality problem? Most people start at the end. That's why so many drop reports blame Google for things the site did, or blame the content for things the calendar did.
1. Clicks or rankings?
Question one. Did clicks fall, or did rankings fall? Different problems. Open the Search Console performance report and switch on all four lines: clicks, impressions, click rate and average position. Read them together, because the combination is the diagnosis.
Here's the reading key. Position down and impressions down: you're ranking lower, a real ranking problem. Position steady, impressions down: fewer people are searching. Position steady, impressions steady, clicks down: something on the results page is taking the click.
One rule before you read anything: compare like with like. Search has a weekly rhythm, so compare equal periods that start on the same weekday, like twenty eight days against the previous twenty eight. Seasonal business? Check the same weeks last year too. A normal week against a holiday week tells you nothing, as you're about to see.
That third pattern has a name now: the AI answer. Pew Research tracked real browsing in March 2025. When a Google search showed an AI summary, people clicked a normal result on eight percent of visits. Without one, fifteen percent. And they clicked a link inside the summary on just one percent.
So if rankings and impressions held but clicks fell on informational queries, check whether those queries now show an AI Overview. Since the end of August 2026, every site in Search Console has a generative AI performance report. It shows impressions in AI Overviews and AI Mode by page. It doesn't show clicks.
Which means a click loss from AI answers isn't a ranking drop, and rewriting the page to rank higher may not fix it. It changes what you measure and which queries you build for. Diagnose it correctly first. Then decide.
2. Real, or a measurement change?
Question two. Is the drop real, or did the ruler change? Search Console isn't a fixed ruler. Google's own debugging guide lists reporting glitches as one of the main shapes of a drop, right next to updates and seasonality.
Two quick checks. Open Google's data anomalies page for Search Console and see if anything is listed for your dates. And remember the newest days in Search Console are fresh data, replaced by final numbers a few days later. A lot of Monday morning panic is just data that hasn't settled yet.
The biggest recent one: around the tenth to twelfth of September 2025, Google stopped honouring a results parameter called num equals one hundred, which rank trackers used to load a hundred results at once. Plenty of those bot loads had been counted as impressions. An agency study of three hundred and nineteen sites found eighty eight percent saw impressions fall afterwards.
It can show up the other way round too. On one site we audited, average position jumped up the scale in four neat daily steps, in the same week as the parameter change, while impressions barely moved. On paper, rankings nearly doubled. Real ranking gains don't arrive in four tidy steps. The deep page two and three impressions that trackers used to trigger simply stopped being counted.
On that site impressions barely moved, so the artefact looked like a too good to be true position. On many other sites it looked like an impressions cliff. Either way, if your comparison crosses mid September 2025, your before and after use two different rulers. Compare clicks instead, or start the window after the change. Verdict on blaming that cliff on lost rankings: placebo.
A habit that saves hours: Search Console added custom annotations in November 2025. Every time Google changes reporting, ships an update, or you deploy something, drop a note on the chart. Six months later, the chart explains itself, and you look like a genius in the meeting.
3. Where is the drop?
Question three. Where is the drop? Sitewide, one section or one template each point somewhere different. In Search Console, compare the two periods on the pages tab and filter by URL pattern: the blog, the product pages, one language folder. The drop is rarely spread evenly.
Read the shape. Every section down by a similar share suggests something sitewide: a holiday, a sitewide technical change, a manual action, or an update that judged the whole site. One folder down points at a change in that section. One template down across thousands of URLs usually means code: a release that touched that template.
Do the same with queries. Split branded searches from non branded ones. If branded queries fell, demand for your name fell, which is a business or season question, not an SEO one. If non branded fell while branded held, look at rankings and the results page.
And split by country. If one market falls and the others hold, look at that market first: a public holiday, a local competitor, a change to the local results page. A site with customers in many countries can barely feel a holiday that halves a site selling in just one.
4. Timing: correlation only
Question four, the one everybody jumps to first. Did it line up with a Google update? Check the Search Status Dashboard, which lists confirmed core and spam updates with start dates and durations. In 2026, the May core update started on the twenty first and ran for almost twelve days. Here's a case from two sites we audited, where the update got the blame for something it didn't do.
Both sites sell to customers in Türkiye. In the last days of May, right in the middle of the core update, their clicks fell hard for six days. Panic would be the normal reaction. Look at average position first.
Position stayed steady on both sites, and impressions fell. That's the demand pattern from our reading key, not a ranking one.
The reason was on the calendar. From the afternoon of the twenty sixth of May to the thirtieth, Türkiye was on holiday for Eid al Adha. Offices shut, families travelled, and fewer people searched for anything work related. The week after the holiday, both sites were getting more clicks than the week before it. Seasonality, not a problem.
That's the trap with timing. Google ships a core update every few months, and each one usually rolls out over one to three weeks. Almost any drop will overlap some update. An overlap is a correlation, nothing more. Before you blame the update, rule out the calendar, the measurement and yourself.
And when it really is an update? Google's own guide says something worth hearing: if a core update moved you, there might not be anything fundamentally wrong with your content. Others may simply have become more relevant. Compare your biggest losing pages with what ranks now, improve for people, and expect the verdict at a later update, not next Tuesday.
5. What did we change?
Question five: what did you change? In our experience, this is where most real drops live. A release that put noindex on a template. A robots file copied over from staging. A rendering change that moved the main content into JavaScript. A firewall rule that started challenging bots.
Search Console holds the evidence for most of these. The page indexing report shows URLs excluded by noindex or blocked by robots. Crawl stats show whether Googlebot's requests started failing. URL Inspection on one affected page shows what Google actually rendered. Then line up the first bad day with your deploy log.
Two reports take ten seconds and rule out the dramatic causes. Manual actions tells you if a human reviewer at Google has acted against the site. Security issues flags hacked content or malware. Google's guide says check both. If either has an entry, that's your answer, and everything else waits.
Bot blocking deserves its own mention, because the site owner often can't see it. CDNs now ship bot controls switched on. Cloudflare, for example, started blocking known AI crawlers by default for new domains in July 2025. That one hits AI assistants, not Google, but the lesson is general: check what your CDN serves to crawlers, not what your browser sees.
A related habit that does nothing: updating the year in every title each January, or bumping dates to look fresh. Google's helpful content guide lists changing dates without substantial changes as a warning sign. Freshness belongs to queries that want it, like a new trailer. Change the content, then change the date. Verdict: placebo.
Only after all five do you ask question six: is it quality? If the drop is real, sustained, not seasonal, not self inflicted, and sits on pages that competitors now beat, you have a content problem. It's the expensive fix, which is exactly why it goes last.
Migrations without disasters
Now the self inflicted drop that costs the most: a migration. New domain, new platform, new URL structure. Almost every migration disaster we've been asked to fix was not a penalty. It was redirects. Missing ones, chained ones, or thousands of old URLs dumped on the homepage.
Google's site move guide is specific. Use permanent server side redirects, 301 or 308. Send each old URL straight to its final destination, not through a chain, even though Googlebot can follow up to ten hops. And don't send many old URLs to one irrelevant page like the homepage, because Google may treat that as a soft 404.
Before launch, the work is a map. Export every old URL that has traffic, links or a place in the index, and map each one to its closest new equivalent. Google suggests testing with URL Inspection for single pages and with scripts for big lists. Changing domains? Submit a Change of Address in Search Console for every verified variant of the old domain.
Why one hop? Every hop adds a request and a delay, and every hop is one more place to break. Chains usually happen by accident: an old migration's redirect points at a URL that a newer migration redirected again. Before launch, crawl the old URL list and flatten every chain, so each old URL lands on a page that returns two hundred.
And the myth that scares people out of doing migrations properly: redirects lose fifteen percent of your authority. In 2016, Google's webmaster trends analyst said plainly that thirty x redirects don't lose PageRank anymore. Google's docs treat a permanent redirect as a strong signal that the target should be canonical. The losses people see come from the mistakes, not the redirect. Verdict: placebo.
Expect some movement anyway. Google says that for medium sized sites it can take a few weeks or more before the new URLs replace the old ones, and longer for big sites. Rankings can wobble meanwhile. That's normal, and it's why the old URL list, the redirect map and a daily check keep running for months, not days.
Back to the patent firm. Here's what we inherited: years of ASP.NET URLs, all dead, plus a WordPress layer with its own set of URLs. We rebuilt the site on a custom Next.js stack and redirected every old URL from both eras to the right new page. Not the homepage. The right page.
Then the rest of the basics. We restored the English pages, fixed hreflang and structured data, and took mobile Lighthouse from the sixties to over ninety, with the main image loading in under two seconds on the landing pages.
The result: from about 400 organic visitors a month to nearly 4,000, in under three months. Plus 830 percent. No secret trick in there. Most of it is a site finally getting back the address book it lost, every old link and bookmark pointing somewhere real again. The same discipline let a multilingual dental clinic we work with move eleven languages to clean URLs and hold its traffic right through the switch.
Wrap up
Here's the episode on one screen. Most drops have a boring cause: the calendar, a reporting change or a release. The update is the last suspect, not the first. And a migration done with a complete redirect map is a growth project, not a gamble.
The thirty minute traffic drop checklist, with the reading key and the migration checklist, is free on the episode page at irongoo.com. Print it before you need it, because nobody reads a checklist calmly on the day traffic falls.
