Google algorithm updates change how its systems assess and order search results. A traffic drop near an update is a reason to investigate, not proof the update caused it. Before you respond, check the confirmed rollout, the pages and queries that lost traffic and other explanations. But fix anything that's broken straight away.
Which kind of Google change are you dealing with?
Five different events get called "an update". Each one needs a different response. Work out which one you might be looking at before you change anything.
Change | What it means | Your response |
|---|---|---|
Core update | A broad change to how Google's systems assess content that doesn't target specific sites | Assess sustained changes across important pages and queries once the rollout finishes |
Spam update | An improvement to the automated systems Google uses to detect spam | Check what the site actually does against Google's current spam policies |
Manual action | A human reviewer at Google found pages that break its spam policies, and Search Console reports it | Fix every affected page, then request a review |
Technical or serving incident | A fault on your site, or in Google's crawling, indexing or serving, changes access or reporting | Confirm which system failed and repair the actual fault |
Discover or search appearance change | One surface, feature or market changes separately from web results | Check the right report and the announced scope before you generalise |
The helpful content system no longer runs as an update of its own. Google's ranking systems guide says it became part of the core ranking systems in March 2024. So there's no separate helpful content schedule to watch for.
Most changes are never announced at all. Search is updated thousands of times a year. Google gives that figure in its post on how we update Search to improve your results. Its core updates guidance says smaller core updates aren't announced because they aren't widely noticeable. The dashboard is a record of notable announced events, not of every ranking change.
An update in the same week as your drop is a candidate explanation, not proof. The timing doesn't show that the update caused the loss. An update isn't a penalty. It also reveals nothing about how much any ranking signal weighs. A clean Manual actions report doesn't rule out an algorithmic change either.
These routes overlap. A broken tag and a genuine ranking loss can land in the same month. So don't force one explanation before you've investigated.
Google's spam policies for Google web search set out what spam updates and manual actions target. The list runs from cloaking and doorway abuse to link spam and scaled content abuse. The policies now cover attempts to manipulate generative AI responses in Google Search as well. That's why you should check your site against the current version.
Check the announced scope, too. The record below shows which updates Google announced as global and which started in one market. Even a global update doesn't land on every site the same way. So judge it on your own Australian queries, audience and page groups, not on overseas commentary.
Confirmed Google updates in 2026 and 2025
Google lists its notable ranking updates on the Search Status Dashboard. Each one has an incident page. It records when the rollout started and finished. As of October 2026, the dashboard shows seven ranking announcements in 2026. That's two core updates, four spam updates and one Discover update.
SourceGoogle Search Status Dashboard, checked in October 2026
The table adds when each rollout finished and what Google announced about its reach.
2026 update | Started | Completed | What Google announced |
|---|---|---|---|
February 2026 Discover update | 5 February | 27 February | Discover only, for English language users in the US at launch |
March 2026 spam update | 24 March | 25 March | Global, all languages |
March 2026 core update | 27 March | 8 April | Core update, rollout of up to 2 weeks |
May 2026 core update | 21 May | 2 June | Core update, rollout of up to 2 weeks |
June 2026 spam update | 24 June | 26 June | Global, all languages |
August 2026 spam update | 18 August | 21 August | Global, all languages |
September 2026 spam update | 24 September | 8 October | Global, all languages, rollout of up to 2 weeks |
Dates are Google's, in US Pacific time. Australia is many hours ahead. So a rollout Google dates to one day may already have started on the next day here.
The latest event at our October 2026 check is the September 2026 spam update. Google says it was complete as of 8 October 2026. That's a dated finding, not a promise that nothing else is rolling out.
The February 2026 Discover update launched for English language users in the US. Google said it would look to expand it to all countries and languages in the future. The update changed Discover, not web search. So it says nothing about Australian web rankings.
2025 update | Rollout dates |
|---|---|
March 2025 core update | 13 to 27 March 2025 |
June 2025 core update | 30 June to 17 July 2025 |
August 2025 spam update | 26 August to 22 September 2025 |
December 2025 core update | 11 to 29 December 2025 |
These dates tell you when Google changed something and how widely it said the change applied. They can't tell you who gained or lost, whether an update caused your drop, or what Google rewarded. A month in an update's name isn't always the month it finished, either. The March 2026 core update completed on 8 April. Treat any list of winners and losers as someone's reading of their own data, not Google's record.
We check the dashboard before every review of this page. We also check after each relevant announcement. For each event we record the incident link, announced scope, start, completion and the date we checked. An expected completion date only becomes confirmed when Google's incident page says the rollout is complete.
How do you diagnose a traffic drop?
Work through five steps in order and write down what you find at each. The early steps are cheap. They also rule out explanations that would make the later steps pointless.
Urgent faults don't wait for the last step. Fix any of these now:
a site that's down
an accidental noindex on priority pages
a security issue reported in Search Console
an enquiry form that's stopped delivering
Google's full week wait is about assessing a core update after it finishes. It isn't advice to leave a broken site broken.
Confirm what actually fell
Write down exactly what you're looking at:
the Search Console property and hostname
the analytics channel
the search type, country and device
the dates and pages
Search Console counts clicks and impressions on Google. Google's page What are impressions, position, and clicks? defines both. Your analytics counts sessions on your site instead. The two won't match. A change in one doesn't prove a change in the other.
Evidence | What it counts | Before you compare |
|---|---|---|
Search Console clicks and impressions | Clicks from Google results and views of your links in them | Same property, search type, filters and complete days |
Analytics sessions | Visits your tag recorded, after consent choices | Tag, consent and channel changes noted on your timeline |
Qualified enquiries | Real leads that meet your written definition | Delivery tested, spam and duplicates removed, the same time to qualify |
Revenue or contribution | Sales and margin from those leads | A window long enough for sales to complete |
Check that the data is complete. The newest Performance report data can be preliminary and still changing, as Google's Performance report overview explains. So compare complete days only.
Search Console's own reporting changes, too. Google's Data anomalies in Search Console page records a logging error that misreported impressions from 13 May 2025. Fixing it on 27 April 2026 could lower impressions in the Performance report. Clicks weren't affected, only impressions and the metrics built on them: click through rate and average position. That means a comparison spanning 27 April 2026 can show an impressions drop that isn't a ranking loss.
Then test the business end: follow a labelled test enquiry from a priority page to whoever answers it. We ran that test in 100 website audits from May 2023 to September 2026. About 1 in 5 sites with an enquiry form failed it. Failing means the request didn't reach the person who responds within one business hour.
The same audits checked tracking on every site. About 3 in 4 had no working GA4 key event for the main enquiry. Our guide to GA4 for small business covers that setup.
A broken tag can cut reported sessions while search clicks hold steady. A failed form can cut enquiries without touching traffic. Measurement, delivery and ranking problems can all happen at once. So fixing one doesn't clear the others.
Put the loss on a dated event record
List everything that changed around the drop, with dates:
Google rollouts
site releases, migrations and redirects
content edits
stock and promotions
tracking
Check the Search Status Dashboard summary for crawling, indexing and serving incidents as well as ranking updates. The dashboard only lists widespread problems. Google's guide on how to use the Google Search Status Dashboard adds a caveat. An issue may be isolated to your own pages. So a quiet dashboard doesn't clear your site.
Test demand instead of assuming it. Australian businesses have their own cycles: end of financial year in June, Christmas, school holidays and the weather. Google's guide to debugging drops in Google Search traffic suggests two checks. Use Google Trends to see whether interest fell across the web or only for your site. Use the Performance report's 16 month view to spot a drop that comes back every year.
If a core update overlaps the drop, Google's core updates guidance recommends a wait. Allow at least a full week after the rollout completes before analysing your site in Search Console. Then compare that week with a week before the rollout started.
Whatever the cause, compare complete, equal periods with the same mix of weekdays. That means a full week against a full week, or 28 days against 28 days. Never compare a part week with a full one. Search Console's comparison filter adds a Difference column to the table. Weekly grouping evens out the day of the week effect, as Google's advanced filtering and comparison page explains.
Periods: [two complete periods of equal length, with dates]
Search type: [the same in both, usually Web]
Filters: [the same country, device, page or query filters]
Grouping: [weekly]
Biggest losses: [pages and queries with the largest negative Difference]
Separate the losses
A total hides the story. Break the drop out by:
page group (service pages, product pages, articles)
brand and non brand queries
search intent
country and device
search surface
Web, image, video and news results are separate search types in the Performance report. Discover has its own report. So don't blend them into one update claim. Google's traffic drop guide says to click through every tab: queries, pages, countries, devices and search appearance.
Treat average position as context, not a rank. It's averaged across every query your site appeared for. So one keyword can move a long way while the average barely shifts.
Small position changes happen all the time. Google's guide says not to focus too much on absolute position. For a walk through each report, see our Google Search Console guide.
Check the other explanations
Each pattern in the data points to a first check. None of them proves a cause on its own.
Observed pattern | First checks | What it doesn't prove |
|---|---|---|
Analytics down, Search Console similar | Tracking, consent, channel attribution and whether pages load | That Google reduced your visibility |
Impressions and clicks both down | Demand, indexing, query coverage, rankings and page availability | That an update caused it |
Clicks down, impressions similar | Click through rate by query, device and page, the live result and the Generative AI performance report | That AI features took the clicks |
Search visits similar, enquiries down | Form and phone delivery, the offer, intent, lead quality and follow up | That more SEO content will fix it |
One template or section loses visibility | Recent releases, canonical and indexing states, and intent match | That the whole site needs rewriting |
Inspect important URLs with the URL Inspection tool. Then check these reports for anything that lines up with the drop:
Page indexing
Crawl stats
Manual actions
Security issues
In URL Inspection, compare the canonical Google selected with the one you declared. That matters because a canonical is a signal, not an instruction. A release or URL change might line up with the loss. If so, our guide to migrating a website without losing SEO rankings covers the checks. For a confirmed technical fault, work through our SEO audit checklist.
Clicks falling while impressions hold is consistent with AI features or other result features taking clicks. But the pattern doesn't prove it. Google's page on AI features and your website covers traffic from AI Overviews and AI Mode. That traffic is included in the overall search traffic in Search Console. It's reported in the Performance report, within the Web search type.
Separately, Search Console's Generative AI performance report shows impressions (not clicks) from AI Overviews and AI Mode. It breaks them down by page, country, date and device. The report reached all websites worldwide as of 31 August 2026. The same AI features page says no additional requirements are needed to appear. You don't need special schema or AI text files either.
Our guide to getting cited in Google AI search covers the Generative AI performance report. Our guide to whether SEO is dead covers what AI search changes about clicks.
A search appearance change can cut clicks with no ranking change at all. FAQ rich results stopped appearing in Google Search from 7 May 2026, as Google's Search documentation updates record. So pages that showed them can lose FAQ impressions and the clicks that came with them. Our guide to schema markup covers structured data and the rich results it can still support.
When search visits hold and enquiries fall, the break sits between the visit and the reply. Our guide to why your website is not converting works through it.
Compare results and state uncertainty
Then look at the results themselves. Search a small, documented set of important queries from the location and device your customers use. Note what's changed: who ranks, what features sit above the first result and how your own listing reads. Ask how the current results serve the searcher differently.
Don't read a ranking rule into a competitor's longer article, extra images or schema. One page outranking you proves nothing about why.
Here's an example listing, with the question to ask about each part.
When impressions hold and clicks fall, Google's traffic drop guide points to two possible causes. Your title and snippet may not be conveying the page well. Or other sites may have more appealing rich results.
Finish the diagnosis in writing:
the most plausible explanation
the competing ones
the evidence for each
what's still unresolved
the next check
If the evidence is thin, say the cause is uncertain. That's a legitimate result. It beats a confident wrong one.
A worked comparison that changes the decision
Here's an example: a teaching dataset, not client results or an industry benchmark. A business compares two complete 28 day windows in Search Console with the same filters. Both sit outside any core update rollout and after the full week wait. The business's enquiry records for both periods are mature. So each enquiry has had the same time to be qualified.
Measure | Before | After | Change |
|---|---|---|---|
Blog clicks | 1,000 | 600 | Down 40% |
Service page clicks | 1,000 | 950 | Down 5% |
Product page clicks | 500 | 550 | Up 10% |
Total clicks | 2,500 | 2,100 | Down 16% |
Qualified organic enquiries | 60 | 62 | Up 3.3% |
The total fell 16%. The whole net loss sits in the blog. Service pages lost 50 clicks while product pages gained 50, cancelling each other out. Qualified enquiries didn't fall with the traffic. Their rise of two is noise, not proof of an improvement.
Chart data:
| Before | After | |
|---|---|---|
| Blog pages | 1,000 | 600 |
| Service pages | 1,000 | 950 |
| Product pages | 500 | 550 |
SourceTeaching example, not client data
The page groups don't overlap. The enquiries are qualified the same way in both periods, with spam, tests and duplicates removed. Keep clicks and enquiries apart. Don't divide enquiries by Search Console clicks and call the result a conversion rate. The two counts come from different systems and measure different things.
The next move is an investigation of the blog queries, not a rewrite. Work out which of these happened:
demand fell
visibility fell
the results changed shape
intent shifted
the site changed
Before you call the lost articles worthless, check whether they helped people who enquired later. Use your analytics paths, or ask new enquirers how they found you. The data supports prioritising that investigation. It doesn't support a claim that Google removed low quality traffic on purpose. And it gives no reason to touch the service pages.
We'll look at your Search Console and analytics data and test your enquiry path. Then we'll tell you which explanation the evidence supports and what we'd check first.
How should you choose and sequence the response?
Let the evidence set the order, not a fixed recovery recipe. What comes first depends on:
the confirmed fault
the harm to customers
the strength of the evidence
how much the pages matter to the business
This is the order we work in, not a Google formula.
Evidence or need | Action | Completion check |
|---|---|---|
Outage, security fault or blocked priority pages | Restore access, correct the mistake and use the relevant report or a specialist | The fault is repaired and important journeys work |
Broken tracking or enquiry delivery | Repair collection and delivery, and annotate the affected dates | A real test reaches the right person and is recorded once |
Sustained loss on important pages | Review intent, evidence, accuracy, useful detail and the competing results | A scoped improvement answers a documented gap |
Overlapping or obsolete content | Review value, users, links and replacement relevance before changing URLs | Each page has a reasoned keep, improve, merge or retire decision |
Not enough evidence | Keep working pages, improve observation and set a next check | The next decision has a date, an owner and an evidence requirement |
Repair confirmed faults first
The first two rows of the matrix don't wait for an update to finish or a complete diagnosis. Google's traffic drop guide notes that security problems such as malware or phishing can trigger warnings. The warnings show up before people reach your site. So treat anything in the Security issues report as urgent. If the site runs on WordPress, our guide to WordPress security for small business covers the basics.
Speed belongs here only when the evidence points to it. If a release slowed the affected templates, check their field data. Our guide to Core Web Vitals explains how to read it. The fixes are in our guide to why your website is slow.
Improve pages with a real gap
For a sustained loss on important pages, compare the page with what the searcher needs. Compare it with the results Google now shows, too. Google's core updates guidance suggests a deeper self assessment when a site sees a large, sustained drop. The assessment uses the questions in Google's guide to creating helpful, reliable content.
The core updates guidance also warns against quick fixes. One example is removing a page element because someone said it was bad for SEO. The guidance recommends leaving content that already performs well alone.
Google's guidance on helpful content says E-E-A-T itself isn't a specific ranking factor. So there's no score to raise by adding a bio or three case studies. Real evidence that a reader can check is what helps. Our guide to E-E-A-T for small business covers what that looks like.
On a service page, check the elements that matter most:
relevant proof
the process
price context
the questions customers really ask
a clear next step
connected supporting pages
Not every page needs all of them. Any trust evidence has to be real. Our guide to service page SEO covers the page in detail.
Don't buy links, reviews or fake branded searches to recover. Don't expect technical polish or schema alone to reverse an update, either. Genuine reviews and professional recognition are ongoing business work, not a recovery tactic. A provider selling those shortcuts after an update is showing one of the SEO red flags worth knowing.
Merge or retire content properly
Update useful content before you consider removing it. Merge pages only when their intent and user journey genuinely fit together. A page with a genuinely relevant replacement gets a permanent redirect to it. A page with no equivalent can return 404 or 410.
Redirecting many old pages to one irrelevant destination might be treated as a soft 404. Sending them all to the home page is one example. Google's site move guidance explains this. A broad category that doesn't meet the same need is no better for the reader.
A noindex does a different job. It keeps a usable page out of search results. Google has to crawl a page to see its noindex. So a page blocked in robots.txt can't pass one on.
Google's core updates guidance calls deleting content a last resort, for content that can't be salvaged. Low traffic on its own is no reason to remove a page.
Every retired page redirects to the home page
Each retired page redirects to its closest relevant replacement, or returns 404 or 410 when none exists
Our guide to content strategy for small business covers the keep, improve, merge or retire decision.
Common mistakes after a drop
The expensive mistakes are the ones made in a hurry, in the week the traffic chart turns down.
Panic rewrites of pages that still work
You lose what was working and can't tell which change helped
Change only pages with a documented gap, in stages
Deleting or redirecting pages in bulk
Useful pages disappear, and irrelevant redirects can be treated as soft 404s
Make a keep, improve, merge or retire call page by page
Acting on social media rumours
Other sites' results say nothing about yours
Start from Google's dates and your own segmented data
Disavowing links indiscriminately
You can discard links that help and gain nothing
Act on links only for a documented link problem
Rewriting pages in bulk with generative AI
Many pages without added value can fall under Google's scaled content abuse policy
Use AI as a helper, then add real detail, evidence and review
Reporting traffic alone
A traffic chart can't show whether enquiries or sales fell
Report qualified enquiries and sales beside clicks
Release changes in stages
Make a small first release where you can. Before it goes live, write down:
the hypothesis
the pages
the exact edits
the date
the person responsible
what you expect to see
when you'll review it
Test forms, navigation, mobile use and indexability before the release and again after it. Know how you'd roll back a broken template or journey.
Affected pages and metric: [pages, metric and the periods compared]
Explanation and confidence: [most plausible cause and the evidence for it]
Other possible causes: [what remains unresolved]
Change and owner: [specific action and responsible person]
Release and checks: [date, function and measurement checks]
Review and rollback: [evidence window and the trigger to reverse]
A record like this makes the result easier to read. It doesn't turn a before and after comparison into a controlled experiment. That's because demand, competitors and Google's systems keep moving while you wait. A staged release improves traceability, not proof.
What to check in five types of business
The diagnosis is the same everywhere. But what you check first depends on how the business wins customers. Treat these as diagnostic scenarios, not lists of any update's winners or losers.
Context | Useful investigation | Avoid claiming |
|---|---|---|
Trades and local services | Service queries, location page groups, the areas you actually cover, and working call and enquiry paths | That suburb pages or Business Profile activity protect a site from core updates |
Ecommerce | Categories, products and buying guides separately, stock, feed and site consistency, seasonality and checkout changes | That a named update punished ecommerce as a whole |
Health and legal | Current facts, contributor scope, the relevant jurisdiction and suitable evidence | That a disclaimer or testimonial makes the advice suitable |
Professional services | Commercial enquiries against generic information traffic, service fit, process and pricing context | That publishing more expert articles guarantees recovery |
Reviews and comparisons | Original testing, method, disclosures, product availability and whether the comparison still helps | That removing affiliate links alone reverses an update |
For a trade business, start with the service and location pages that bring in calls. Our guide to SEO for tradies covers the local side.
In a store, a stock change can move one page group without the others. So can a mismatch between the product feed and the site. Our guide to ecommerce SEO for Australian online stores covers categories, stock and product data in full.
Health and legal pages need current facts, written or reviewed by someone working within their scope. That detail stays with our guides to SEO for medical practices and SEO for law firms.
For an accountant or consultant, separate commercial enquiries from information traffic. Losing visits to a tax rates explainer is one problem. Losing the service page that brings in clients is a different one. Our guide to SEO for accountants covers the regulated side for accountants and financial advisers.
Protect the business while you investigate
Work out what the drop costs before you react to it. Which lost page groups brought qualified enquiries, completed sales and contribution after delivery costs? Count how long enquiries take to arrive and convert. Check whether the team could handle more follow up.
The same percentage drop can matter a lot to a retailer. For a service firm whose lost pages never sold anything, it can matter very little.
If you run a paid search test, give it:
its own affordable cap
one objective
a landing page that suits the searches
working measurement
Search ads catch demand that already exists. A test can also lose money. Paid search isn't an automatic substitute for lost rankings or a guaranteed bridge back to them. Our guide to whether Google Ads is worth it for a small business works through that decision.
How long does recovery take after a traffic drop?
Repair and reassessment run on different clocks. A technical fault can be fixed today. But crawling, indexing and reporting take time to catch up. Content improvements have no fixed timetable and no promise of a return to an earlier position.
Google's core updates guidance says some changes can take effect in a few days. But its systems can take several months to learn and confirm that a site keeps producing helpful content. You don't always have to wait for the next core update to see the effect. Still, Google says a few months with no effect could mean waiting for the next one. And there's no guarantee your changes will make a noticeable difference.
With two or three core updates a year, waiting for the next one can mean months. That's a reason to make fewer, better changes and record them, not to keep rewriting while you wait.
A manual action has its own route. Google's help page on the Manual actions report explains how it works. A human reviewer at Google issues one for pages that don't comply with its spam policies. You'll see it in that report.
Fix the issue on every affected page, because fixing only some won't earn a partial return. Then select Request Review and explain what you fixed. Google says most reconsideration reviews take several days or weeks. An ordinary core update decline has no equivalent: no request to file and no one to appeal to.
After a spam update, recovery depends on the issue and on genuine compliance with the policies. Google's page on spam updates says sites breaking its policies may rank lower or not appear at all. Where links were the problem, any ranking benefit they generated is lost and can't be regained. So don't plan around getting it back.
Set a review window from the traffic volume, the business cycle and the cause. A page with a few hundred clicks a month needs a longer window than a busy category. At the review, record one of three outcomes: improvement, no clear change or deterioration. Reopen the diagnosis whenever the evidence contradicts your first explanation. Our guide to how long SEO takes to work covers realistic timelines.
How to prepare before the next update
The best time to prepare for a drop is before it happens. Start with a baseline you trust and one person who owns the checks. Combine routine reviews with triggers. A fixed schedule alone misses a changed rule, a withdrawn product, a broken form or a major release.
Workstream | Routine | Trigger for earlier action |
|---|---|---|
Search and enquiry monitoring | Review a manageable set of business critical page groups | A material sustained change or a failed enquiry test |
Priority services and products | Check the process, price context, availability and next steps | An offer, staff, stock or delivery change |
High stakes information | Assign a qualified owner and a review date | A change in the rules, the evidence or professional scope |
Technical changes | Keep a release log and run functional checks | A migration, template change, tracking release or incident |
Update record | Check Google's history before reviews and after announcements | A new confirmed event or a completed rollout |
Keep dated baselines and the names of the people who contribute. Use a client example only with permission. Publish content with a clear customer purpose, then maintain it. Skip fixed article quotas, fake first hand experience and new dates on pages nobody reviewed.
Track visibility and outcomes together. Define a qualified lead once and stick to it. Branded search demand is a useful business signal. But it isn't a confirmed direct ranking factor. For the wider plan this routine sits inside, read our guide to SEO for small business.
Daily rank checks mostly create noise. So review on a steady rhythm and act when a trigger fires. Here's the whole approach in one scorecard.
Bottom line Confirm the problem, make a proportionate change and review comparable evidence
What we recommend at Elev8d
Start with the boring checks. We confirm the drop is real:
equal, complete periods
the right property and search type
a test enquiry that reaches a person
tracking that records it once
Anything broken gets fixed that day, update or no update.
Then we split the loss by page group, query type and surface. We line it up against Google's dates and the site's own release log. We look at the live results too. That narrows the problem to the pages that matter. Where the evidence doesn't settle it, we say the cause is uncertain rather than guess.
Content changes come last, and only where the evidence shows a gap a customer would notice. We make them in small, recorded releases. Pages that still work get left alone. We judge the result on qualified enquiries, over a window long enough to read.
What we won't do is rewrite a site because an update landed the same week. We don't buy links or promise a recovery date. Say a drop has cost real enquiries and you need cover while you investigate. Then a capped paid search test can make sense, as long as it's measured on its own.
FAQ
What if traffic dropped and Google hasn't announced anything?
Run the same diagnosis with the update off the list. Your release log, tracking, demand and the live results now carry more of the weight. An unannounced change is still possible but can't be confirmed. So don't build a plan on it.
What if the comparison is inconclusive?
Leave the working pages alone. Instead, improve what you can observe. That means cleaner page groups, a regular test enquiry and a note against every release. A longer window can settle it. If it still doesn't, make one small, reversible change where the evidence is strongest and record it.
Why haven't my rankings come back after the fix?
A repair restores the page, not its position. Rankings are relative. Google's core updates guidance says other pages may be doing a better job of helping the searcher. Compare your page with the ones now above it. Improve it if the gap is real.
Next steps
Check one important page group
Pick the page group that brings most of your qualified enquiries. Then compare two complete, equal periods in Search Console with the same filters.
Confirm one explanation
Test the most likely explanation against Google's dates, your release log and a labelled test enquiry. Then write down what you found.
Assign one useful action
Give one person one action with a date: a repair, a scoped improvement or the next evidence check. If nobody in house can take it, hand it to our SEO team.