There’s a certain type of Monday morning I’ve come to recognise. Client emails landing before 9am. Subject lines like “our rankings have dropped?” with a question mark, as if they’re not entirely sure whether to be worried yet. Screenshots of rank trackers that look like a seismograph reading.
Last week was one of those weeks — and for good reason. Google dropped a spam update and a core update within 72 hours of each other, creating the kind of SERP volatility that sends the SEO industry into a collective spin. Add in a new AI crawler that ignores your robots.txt, and AI Overviews quietly eating more of your organic traffic, and you’ve got a week worth actually paying attention to.
Here’s what happened, what it means, and what I’d actually do about it.
The Double Update: One to Shrug At, One to Watch
The spam update landed on 24th March and was done in under 20 hours — apparently the fastest spam update Google has ever rolled out. The core update followed on 27th March and is still completing as I write this, with up to two weeks expected for full rollout. SEMrush had volatility scores hitting 9.5 out of 10. Over half of tracked domains saw ranking movement.
Here’s the thing though: for most well-run client sites, the spam update will have passed without a scratch. Spam updates are designed to clean up the index — targeting manipulative link profiles, scaled junk content, and sites gaming the system. If your clients aren’t running PBNs or publishing auto-generated rubbish, there’s not much to worry about there.
The core update is a different conversation. Core updates are a reassessment of quality across the board — not a penalty, but a re-ranking. Google looks at your pages relative to competitors and decides whether you still deserve the spot. Some sites go up. Some go down. Neither means you’ve done something wrong.
What I’d actually do: Check Search Console for any manual actions first — that’s how you confirm whether the spam update touched you. If there are none, and your drops align with the core update timeline (27th March onwards), that’s a content quality signal, not a penalty. Don’t make reactive changes mid-rollout. Wait until mid-April when the dust settles, then look at which pages dropped and what outranked them.
The worst thing you can do is start rewriting pages while the update is still processing. I’ve seen that make things worse. Sit tight, pull the data, then act.
AI Content Without Real Expertise Is Now a Liability
The early patterns from this core update are telling a consistent story: AI-assisted content with genuine human input is holding up fine. AI-generated content that’s fluent but hollow — where the AI is doing all the thinking — is dropping.
This doesn’t surprise me. I’ve said for the last two years that AI is a tool, not a strategy. Using it to bash out 40 blog posts a month for a local business, with no real editorial oversight and nothing original to say, was always going to have a shelf life. That shelf life appears to be ending.
The nuance here is important. Google isn’t penalising AI use. It’s penalising content that fails to demonstrate original thought, real expertise, or anything the user couldn’t get from reading the top three results that already exist. If the only thing your content does is repackage what’s already out there, it’s going to struggle. Increasingly, it already is.
What I’d actually do: Go through any client content that was produced at scale over the last 12 months. Ask one question about each page: does this add anything? Real examples, genuine opinions, client-specific data, local context — anything that proves a human with relevant knowledge was actually involved. If the answer’s no, either improve it properly or consolidate it. Thin page farms are a liability now, not an asset.
Author signals matter too. Real names, relevant credentials, and internal links to supporting content all help. It doesn’t have to be complicated — but it has to be real.
AI Overviews Are Eating Informational Traffic — and Growing Fast
This is the bigger structural story, and it’s one I’m having with clients almost every week at the moment.
AI Overviews now appear in roughly a quarter of all US Google searches, and that number is climbing. When an AI Overview is present, organic click-through rates drop dramatically — studies are consistently putting it at less than half of what you’d expect without one. For informational content especially, the economics of SEO are genuinely shifting.
The thing that often gets missed in these conversations: AI Overviews appear far less frequently on transactional and local searches. A query like “what is a debenture” is almost certainly going to trigger one. A query like “conveyancing solicitor Sheffield” or “emergency boiler repair Leeds” largely won’t. For the majority of the clients I work with — local businesses, tradespeople, professional services — the direct impact is more limited than the headlines suggest.
Where it does matter: the “brand citation” effect. Brands that appear as cited sources inside AI Overviews see a meaningful uplift in clicks compared to those not cited. That’s the new game for informational content — not just ranking, but being cited.
What I’d actually do: Two things. First, reframe client reporting now, before the next traffic dip conversation becomes uncomfortable. If your reports lead with sessions and organic traffic, start adding leads, enquiries, and conversions alongside them. Traffic going down 15% while leads stay flat is a fine result — but only if you can show that. Second, for clients who rely heavily on informational blog content, start structuring articles to be citation-friendly: short summary paragraphs, direct answers up top, clear definitions. That’s what gets pulled into AI Overviews.
Google-Agent: The AI Crawler That Doesn’t Care About Your robots.txt
On 20th March, Google quietly added something to its official crawler documentation that I think a lot of agencies haven’t clocked yet. A new entity called Google-Agent — a user-triggered AI fetcher that, unlike Googlebot, completely bypasses robots.txt.
It’s used by Google’s AI systems, including a product called Project Mariner, which lets users direct an AI agent to browse the web on their behalf. Because it’s acting on a user’s instruction, Google treats it more like a browser than a crawler — and robots.txt simply doesn’t apply.
To be clear: for the average SME client running a standard business website, the immediate risk here is low. Google-Agent isn’t mass-crawling anything. It shows up when a user actively points it at a URL.
But if you manage client sites with staging environments that are accidentally internet-accessible, or content restricted behind a robots.txt block rather than actual authentication, that’s worth a look. Robots.txt was designed to manage crawlers. It was never the right tool for securing content — and now there’s a formally documented agent that will walk straight past it.
What I’d actually do: Run a quick audit on robots.txt files for any client with sensitive or restricted content. If something genuinely needs to be protected, it needs server-side authentication, not a disallow rule. For most clients, no action needed — but file it away as context, because Google-Agent will only become more widely used as AI tools expand.
The Practical Summary
Last week was noisy. But the signal underneath it is the same message that’s been building for a while: Google is getting better at rewarding sites that are genuinely useful, and filtering out everything else. The double update didn’t change the rules — it just enforced them more forcefully.
For agencies, the priorities coming out of this week are straightforward. Wait for the core update to stabilise before drawing conclusions. Audit client content for anything thin or AI-generated without real editorial value. Start shifting how you report results — traffic is one metric, not the metric. And get ahead of the AI Overview conversation before a client raises it as a problem.
Practical SEO — solid on-page, real content, legitimate links — still works. It’s just that Google is now a lot less forgiving of the shortcuts.
