🤖

​Google launches AI Mode in the US. AI Mode uses Gemini to offer a conversational style interface, similar to ChatGPT Search or Perplexity. According to Google, this is "only a glimpse of what's to come."

📉

​Clicks from AI Mode are currently not trackable in GSC but will be soon. It was initially believed that clicks from AI Mode would not be trackable at all but John Mueller has confirmed this as a bug they are trying to fix.

🗃️

​Deep Search adds new research capabilities to AI Mode. Deep Search can conduct hundreds of searches at once, from multiple sources and formats to produce a cited report-style result in a few minutes.

📷

​Live Search allows multimodal searching in AI Mode and Google Lens. This new feature uses Project Astra's technology to answer real-time questions about what a user's camera is seeing.

🎟️

​Agentic search will soon allow AI Mode to complete purchases on your behalf. AI Mode will be able to search for, compare and purchase things based on criteria set by the user. This will begin rolling out with event tickets.

🛍️

​Shopping in AI Mode includes price tracking, virtual try-ons and agentic checkouts. AI Mode can provide recommendations, compare prices across multiple shops and allow users to virtually try-on products in Google search.

🙋

​AI Mode will bring personalisation directly to search results. Users can opt-in to personalised search based on past searches and usage of connected Google apps such as Gmail. This could be the end of SERPs as we know them.

📊

​AI Mode can create custom charts, graphs and visualisations directly in search. This feature can summarise multiple datasets into unique visualisations (i.e. not pulled from the sources) and will begin with sports and finance queries.

🤑

​Ads have been confirmed for both AI Overviews and AI Mode. This had been speculated for a long time and Google confirmed that ads are available in AIOs for US users and will be coming to AI Mode soon.

Get ready, London! 🚀 The biggest Search ‘n Stuff Signature London Meetup of the Year is hitting the city this summer, and trust us — this isn’t just another networking event with stale sandwiches and awkward small talk.
Nope. This is the rooftop party you’ll be talking about all year.

📅11 June 2025
​📍Doggett's Coat and Badge, London
🍔 Full BBQ feast – including vegetarian and vegan options (we see you, herbivores)
🍹 4–5 drinks per person – enough to fuel bold conversations and legendary stories (alcoholic & non-alcoholic options available)
🎤 4 Quickfire Talks – topics covering SEO, Strategy, PPC, and Paid Social (Speakers to be revealed VERY soon 👀)
🎥 Professional video and photo coverage – because if you don’t post about it, did it even happen?
🤝 Unbeatable networking – meet a fantastic crowd of 60+ marketers, founders, agencies, freelancers, and in-house pros from all across the industry

Google can tell if you're the correct English

Google detects website language automatically, it doesn't use hreflang for that. It's especially important when you're dealing with things like UK vs US websites where the English language is localised. 🌍
​
One of the most common international SEO pain points I see is US pages ranking the UK SERPs and vice-versa, despite hreflang being "correct".
​
Localisation is important (aluminium vs "aluminum" anyone?) and one easy way to audit this is the Screaming Frog Spelling & Grammar check. 🐸
​
In this instance, I setup a filter just to include the /us/ of my website, and specifically told it to use the US English dictionary.
​
You'll get a helpful report of where you've used correct English (i.e. British English), where you should be using American English. 📘

Don't build your strategy based on cosine similarity

I have some facts and some concerns about the latest SEO obsessions of turning your content into a big load of numbers to do stuff with that I want to share with you:
​
1. If you're being sold that this kind of approach is the key to visibility in things like AI Overviews, that is just incorrect. Google at the moment is leaning incredibly hard into a bunch of site-wide metrics they have. I can pretty much write whatever I like on LinkedIn Pulse and have it appear in an AIO, in a Featured Snippet and organically in a matter of hours (Lily Ray is queen of these examples, check that out). The context of the site far outweighs whatever percentage point of "semantic relevance" you think you're engineering.
​
2. Machines turn content into numbers because it's a helpful way for machines to process it, because they can't see it the same (flawed) way humans can. The goal of the machine is to try and build systems to understand what humans like. If you become obsessed with machine metrics, you may miss the larger picture of just making cool stuff that people like - which the machines are geared to gravitate towards.
​
With that said - there are lots of useful and cool things you can do with embeddings and understanding how machines view content; arguably that's going to be even more important in an agentic future.
​
❌ Search (even if you're talking "AI") is still nuanced, multi-faceted and these "new" approaches are not silver bullets.
​
❌ Don't let the tactic or tool destroy the strategy.
​
❌ When the metric becomes the goal, it becomes a bad metric
​
On Sunday I wrote a list of "top SEO experts" on LinkedIn Pulse. 1 day later it ranked #1 in the Featured Snippet and #1 organically. Today it has started appearing in AIOs.

We get the impression Google likes AI...

Some initial thoughts from Google I/O, apart from the KilledByGoogle Bingo card being filled in the first 2 minutes:
​
✨ AI Mode: Zero surprises here. Google has to compete with other AI-powered conversational search, but it's not going primetime until they can figure out how to monetise it. I briefly spoke to Mordy who also saw this coming 17Ly away, this is the future of search.
​
✨ Strategically, the same SEO game plan of trying to be as helpful, relevant and trustworthy to your users/customers shouldn't change. Just like the early days of SEO, there are so many 'tricks' that work with AIOs/AI search at the moment. People will try and tempt you with them to make a quick buck. It will work until it doesn't.
​
✨ Right now, ranking #1 for a broad key phrase means you get a lot of clicks, but many people will drop out when the price is wrong, the brand isn't a good fit for them, or doesn't fit some kind of personal requirement. The personalisation we'll see within AI-type systems will mean a "broad" term with 50,000 searches a month, may generate 1,000 different set of results/recommendations based on the user profile. Fewer clicks, but much higher qualification are likely the future - so as Myriam Jessier has been publishing, making sure AI systems understand what your brand is, will be paramount.
​
✨ Google is eating informational content queries and people saying "it doesn't matter, they don't drive revenue" are truly myopic. Being visible 100,000s of times a year to people and helping them is part of how you build a brand. Google wants you to forget those brands, and become reliant on coming to them for answers - this is absolutely key for them building revenue. There will be a long-term deterioration of brand recall built through search, as Google closes that net.
​
✨ Hardware and eco-system owners; Google, Apple, Microsoft, will win the AI-search game. They are the ones that have all of the surrounding user data that will allow the level of personalisation to happen that will make agents actually helpful.
​
✨ Google is pushing the narrative that AIOs push more clicks to other websites. As Lily Ray points out, this is some very careful wording, with generative results sometimes listing dozens of sites, it absolutely reduces the amount of clicks you are getting for being "#1" (which AIOs are counted as, no matter where you are), so don't expect GSC clarity on that click data any time soon.
​
✨ Longer term, I expect there will be some legislation around AI generation because Britney Muller points out, Google are no longer just surfacing the world's information, they are interpreting, and Barry Adams 📰 suggests now makes them actually a publisher - so there should be some legal recourse for the content they produce. Google is orders of magnitude larger than ChatGPT, so the "bad things" that can happen with iffy AI hallunications are going to increase hundred fold over the next couple of years.

Do what's best for business

"Do what is best for your business, in the most SEO-friendly way possible" is a better approach than deciding what to do based on SEO. This approach is the one that survives every "SEO is dead" litany, even the the latest AIO/AI Mode hype.
​
I think it's some of the most interesting changes we've seen in almost two decades, there is an enormous opportunity ahead, and it's obvious to me that a lot of consumer thinking, consideration and decision making is going to be outsourced to agents.
​
They'll be new tools, new processes, new players, new winners, new losers, but the core tenants of our work remains unchanged.
​
There are no "LLM/GEO/AI optimisation experts" yet, there's a lot up in the air, and a lot of the things we saw at Google I/O are still quite far off primetime, so take a breath 😇
​
​I've recorded a Search with Candour episode with Jack Chambers-Ward to chat through all this!​

Understand the intent proximity of questions

When it comes to AI Mode, Gemini is going to branch out with close proximity questions to build an answer, so it's important we understand what they are.

​AlsoAsked is currently a great way to do this - and if we get more data from AI models, we will add it in 💙

Try it: https://alsoasked.com​

Top Core Updates referrer leaderboard

A big thank you to our top referrers, who have signed up over 10 people to the Core Updates newsletter, go follow them!

🥇 MJ Cachón Yáñez (LinkedIn / Bluesky)

🥈 Nikki Pilkington (LinkedIn / Bluesky)

🥉 Lidia Infante (LinkedIn / Bluesky)

🏅 Saijo George (LinkedIn / Bluesky)