The core update ends, a year of bad impressions data surfaces, and WordPress 7.0 slips.

A dense week. Between 6 and 12 April, Google closed out the March 2026 core update after nearly two weeks of churn, then quietly told everyone their Search Console impression data has been wrong since May 2025. Meanwhile AI Mode crept deeper into the standard results page, Chrome shipped a genuinely useful CSS accessibility feature to every major engine, and WordPress 7.0 blew through its 9 April release date. If your business depends on being found and chosen online, here's what actually happened — and what we'd do about each item.

1. The March 2026 core update is done — now the real reading starts

What happened

Google confirmed on Wednesday 8 April that the March 2026 broad core update had completed rolling out — 12 days and 4 hours after it began on 27 March. Google called it "a regular update" and published no companion post, but the tracking community saw meaningful movement: some analyses suggested a large share of domains saw at least some ranking change, with content quality, originality and authority the recurring themes among winners.

The post-update review that's actually worth doing

• Wait for the completion date, then compare. Now that 8 April is the official end, compare the two weeks after it against the two weeks before 27 March. Mid-rollout data is noise; post-rollout data is signal.
• Segment by page type. Core updates rarely hit a site uniformly. Losing blog rankings while service pages hold tells a very different story from the reverse.
• Judge revenue pages first. A visibility dip on informational posts matters far less than movement on the pages that book calls and take orders.
• If you gained, bank it. Winners should double down on what the update rewarded — genuine expertise, original detail, clean structure — not coast.

2. Search Console's confession: impressions were inflated for almost a year

What happened

Buried in the data anomalies report, Google disclosed a logging error that inflated impression counts in Search Console going back to May 2025. Almost a full year of impression data was overstated. Then, on Sunday 12 April, many Search Console users received emails claiming Google had "started collecting Search impressions" for their properties — emails Google later admitted were sent in error. Two data wobbles in one week, on the platform most businesses treat as ground truth.

What this means for your reporting

• Your visibility didn't drop — your ruler changed. If impressions fall off a cliff in your April reports, check whether clicks fell too. Corrected counting is not lost traffic.
• Re-baseline anything built on impressions. Year-on-year impression comparisons that straddle May 2025 are now unreliable. Annotate dashboards so future-you doesn't misread the step change.
• Weight your KPIs toward outcomes. Clicks, enquiries, orders and revenue were never affected. This week is a strong argument for reporting on what the website produces, not what a third-party counter says it was shown for.

3. AI search kept expanding — and got fact-checked

Three connected AI-search stories landed this week. First, Google was spotted testing the AI Overviews "Show more" jump straight into AI Mode on desktop — extending the mobile behaviour announced in January, and pulling more query journeys out of the classic results page. Second, Search Console appears to be testing a new "AI contribution" report, a hint that AI-driven visibility is about to become officially measurable. Third, a New York Times–commissioned study found accuracy and grounding holes in AI Overviews, and Sundar Pichai used a podcast appearance to sketch an agentic, "many threads running" future for Search.

How to respond without chasing hype

• Assume more zero-click journeys. As AI Mode absorbs follow-up questions, the visits you do get arrive later in the decision, deeper in your site. Every service page needs proof and a clear next step.
• Be the quotable source. AI answers with accuracy problems lean harder on well-structured, verifiable pages. Clear headings, direct answers and real authorship are the inputs — all decided during html web development, not bolted on afterwards.
• Watch for the AI contribution report. If it ships, baseline it immediately, the same way smart teams baselined AI referral traffic early.

4. Platform week: contrast-color() goes Baseline, Chrome 148 hits beta

Quietly excellent week for the web platform. Chrome 147 reached stable, shipping the CSS contrast-color() function — which, now present in all major engines, became Baseline Newly available. It takes any colour and returns black or white, whichever contrasts harder: a one-line accessibility win for buttons, badges and theming systems. On 8 April, Chrome 148 went to beta on Android with Web Serial API support and re-enabled SharedWorkers, and the 148 cycle also brings native lazy loading for video and audio elements.

In the same week, Google's Search Off The Record podcast dug into page weight — how ballooning page sizes hurt users and crawling — following updated documentation clarifying Googlebot's file size limits. The message to builders is consistent: heavy pages cost you twice, once with visitors and once with the crawler.

5. WordPress 7.0 misses its 9 April release date

WordPress 7.0 did not ship on 9 April as scheduled. The core team announced in late March that the release needed more time in the Release Candidate phase to stabilise the data storage layer behind its headline feature — real-time collaboration — and this week the slip became concrete: pre-release versions are paused through 17 April, with a revised schedule promised by 22 April. The 7.0 cycle still targets collaboration, AI infrastructure, an admin refresh and a higher PHP minimum.

• If you run WordPress: do nothing dramatic. Stay current on the stable branch, keep plugins and themes updated, and let 7.0 land when it's ready.
• If you're mid-project: don't build against 7.0 assumptions. Ship on today's stable and plan the upgrade as a separate, tested step.
• If you're choosing a platform: the delay is actually a good sign — shipping a data-layer feature before it's stable would be the real risk. The same logic applies on commerce builds: serious Shopify development partners stage risky changes behind testing rather than launching to a calendar.

6. "Mt. AI": Grokipedia keeps falling, and spammy links get ignored

Two quality-signal stories rounded out the week. Grokipedia's search visibility continued its heavy slide — the now-familiar pattern where sites scaling hard with AI-generated content surge in Google, then drop as quality systems catch up. And Google's John Mueller reconfirmed that Google may simply ignore outbound links from sites that violate its spam policies — a reminder that links bought from junk sources often aren't penalised, just silently valueless.

Both stories point the same direction: shortcuts decay. Volume content and cheap links keep producing the same chart — a spike, then a cliff. The durable curve belongs to sites with fewer, better pages and links earned from work worth citing.

What this week means for your website strategy

Pull the threads together and three decisions emerge for anyone who owns a business website or is scoping a redesign:

• Measure outcomes, not exposure. A year of inflated impressions broke a thousand dashboards but zero businesses. Build reporting around enquiries and orders, and no logging error can lie to you.
• Build for the machine-readable web. AI Mode on desktop, an AI contribution report, accuracy-checked AI Overviews — every signal says structure, speed and verifiability are the new distribution layer. Those are build-time decisions. Ask any agency you're evaluating — whether that's a local shop from a "web developers near me" search or a specialised remote team — to show how their builds earn visibility, not just how they look.
• Respect the boring disciplines. WordPress delayed a flagship release to protect data integrity; Google spent a podcast on page weight. The industry's most sophisticated players are obsessing over fundamentals. So should your website — and your shopify developer partner, if commerce is your engine.

Frequently asked questions

When did the March 2026 Google core update finish rolling out?

Google confirmed completion on 8 April 2026, roughly 12 days after the 27 March start. With the rollout closed, you can finally read Search Console as post-update reality — compare the fortnight after 8 April against the fortnight before 27 March, and segment by page type before drawing conclusions.

Why are my Search Console impressions suddenly lower in April 2026?

Google disclosed a logging error that inflated impressions going back to May 2025. Corrected counting looks like a drop but isn't one — check whether clicks moved before worrying, and annotate your dashboards so the step change doesn't mislead future reports.

Should I have panicked when WordPress 7.0 missed its 9 April date?

No. The delay protects the data layer behind real-time collaboration — pre-releases are paused through 17 April with a new schedule due by 22 April. Stay on stable, keep everything updated, and treat the eventual upgrade as its own tested project.

What did Chrome ship for developers this week?

Chrome 147 hit stable with CSS contrast-color(), now Baseline across all major engines, and Chrome 148 reached beta on Android on 8 April with Web Serial and SharedWorker support, plus native lazy loading for video and audio in the 148 cycle. The contrast-color() function is the one to adopt first — it's a one-line accessibility upgrade.

Does scaling content with AI still work after this core update?

The Grokipedia slide tracked this week is the pattern: AI-scaled content surges, then drops hard once Google's quality systems catch up. The core update rewarded originality and authority. Fewer, genuinely useful pages beat volume — every time the chart runs long enough.

The takeaway

This week's theme was trust in the data — and in the fundamentals. Google corrected a year of its own numbers, a major study corrected AI Overviews, and WordPress chose correctness over a calendar. The websites that came through untouched were the ones that never depended on inflated metrics or scaled shortcuts in the first place: fast, structured, genuinely useful, and measured on revenue rather than impressions.