It’s been a busy week in the world of SEO news. If you’re publishing content, then this is the stuff that you need to be aware of.
This week’s TL;DR
There have been a few key things to emerge from the world of search this week – and Google finally letting publishers opt out of their content being used to create AI Overviews will see many rejoicing. Finally, ranking on Google won’t come at the cost of, well, supporting a feature that can stop people clicking on your site.
Sticking with Google, Personal Intelligence has rolled out, allowing you to connect apps like Gmail and Maps to personalise the experience. If it’s useful, it could radically change search, but it looks like the finer details aren’t quite as good as many had hoped.
Like many that work in the world of search, we’re constantly wondering how LLMs are going to shake up the industry. There are a few tools out there that promise to track any mentions of your brand, but research by Rand Fishkin (from SparkToro) suggests that there’s around a 1% chance of chatbots returning the same answer twice… highlighting how the ways of tracking search aren’t as useful when it comes to LLMs.
All this and more in today’s newsletter… read on to find out everything you need to know about in the world of search.
Would you opt-out of AI Overviews?
In the same week as Google pushes a new follow-up feature to its AI Overviews, the UK government has called on the search giant to allow sites to ‘opt-out’ of appearing in the AI generated result, without impacting organic search rankings.
Google responded swiftly, noting it wants “to protect the helpfulness of Search for people who want information quickly, while also giving websites the right tools to manage their content.” It also highlighted how it had “[launched] Google-Extended, a new* control that lets websites manage how their content is used to train our Gemini models.”
The consultation continues.
(*Google calling this control ‘new’ is interesting, given it appeared nearly three years ago.)
the better web co. insight
This is a key move from the UK government, and is being met with cautious optimism among publishers. While the consultation is designed to give “stronger bargaining power and support the long-term sustainability of trusted information” it remains to be seen how Google will push back and whether improved citations will do anything to offset the loss of clicks from AIOs.
Being able to opt out of AIOs is fine, but it doesn’t stop them from appearing, and will potentially stop people clicking on your site – unless your content is 100% unique and can’t be found anywhere, it’s going to take a strong stance from all publishers to see if being able to remove your site from AIOs makes any difference.
Read more
- Google Blog: Just ask anything: a seamless new Search experience
- Gov.uk: CMA proposes package of measures to improve Google search services in UK
- Google Blog: Our approach to website controls for Search AI features
Keep building trust
Only around one third of Americans have a high level of trust in AI search results, according to a recent Statista Consumer Insights survey.
Interestingly, there was slightly more trust from Americans when it came to searching health and wellness topics (35% having a high level of trust), while trust was lower for finance and news results (29% and 27% respectively having a high level of trust).
the better web co. insight
It’s important to see the nuance here, as these tests were only looking at trust in YMYL topics, i.e. heath, finance (with news also surveyed). Looking at the results, it’s quite polarising, with those that completely distrust AI the largest group… but then followed closely by those that have full trust in the source.
It’s worth noting that this survey was conducted in the US, a country that’s far more cautious over the use of AI compared to places like the UAE, Norway or France – it will be interesting to see how this changes in the next year if the survey is run again.
Read more
- Statista: Limited trust in AI Search result
- Microsoft blog: Global AI adoption in 2025—a widening digital divide
What’s your video strategy?
If you’re looking to increase visibility and citation source presence in LLM results, you might want to take another look at your video strategy.
According to Adweek, YouTube has overtaken Reddit as the leading social citation source, with LLMs now better able to understand video content thanks to a proliferation of transcripts, explainers and deeper descriptions.
That said, content such as YouTube still only accounts for a very small portion of LLM citations, with a Spotlight study that analyzed over 1.2 million citations from eight LLMs seeing YouTube only make up a small percentage of references. Written content made up the majority of the sources, with guides, tutorials, articles and reviews prominent.
However, a separate study from Ahrefs also showed that if you’re only interested in brand mentions on LLMs, then appearing in a YouTube video is one of the best ways to get that visibility.
the better web co. insight
This feels like LLMs are starting to catch up with traditional SEO ranking guidance, where having rich media beyond the written word is key. It makes sense that Reddit would have been a huge source for AI, given it’s the perfect combination of insight and relevance to a lot of queries. How much content has been already used in the training is up for debate, but it’s clear a strategy that gets your organically mentioned in YouTube videos is becoming increasingly powerful.
Read more
- Adweek: YouTube overtakes Reddit as ho-to citation source on AI Search
- Spotlight: What content types do LLMs prefer?
- Ahrefs: Top brand visibility factors in ChatGPT, AI Mode, and AI Overviews
AIs do not give consistent recommendations
Are you tracking AI visibility? New research suggests that might not be as straightforward as some of the AI tracking tools claim.
Rand Fishkin, of SparkToro created an extensive experiment with 600 volunteers who ran the same AI prompts multiple times and recorded the responses.
“The bottom line is: AIs do not give consistent lists of brand or product recommendations” Fishkin notes, as the results of his experiment shows “there’s a <1 in 100 chance that ChatGPT or Google’s AI, if asked 100X, will give you the same list of brands in any two responses. Claude is just slightly more likely to give you the same list twice in a hundred runs.”
the better web co. insight
This test shows how different AI search is to Google ranking – the huge variety of responses is far more reminiscent of asking a group of different people for their thoughts on an issue each time you have a question. In a way, that’s good as it puts the decision back in the hands of the searcher – getting more varied information before making a decision. However, the thought that you might need to ask five times (or more) to make an informed choice would put off a lot of people. It’s important to know how brands are ranking in LLMs, but this data suggests there are no clear ‘rules’ on how a brand can consistently appear in AI chatbots.
Read more
What happens if there’s no AI Overview?
If you’ve spotted an odd result at the top of SERPs which isn’t obviously an AI Overview or featured snippet, you’re not alone. As Epi founder, Alex Greenland, flagged on X it appears Google is trialing a new search response format.
Ragan Patel, Google’s VP for Engineering for Search, has responded to confirm this is a fallback when an AI Overview can’t be generated. In this case, the search engine displays a featured snippet – though he admits it could be labelled better. The good news is he’s asked the team to improve the formatting.
the better web co. insight
It will be interested to see how this will be received by users – while featured snippets were a precursor to AI Overviews, as they’re more like an expanded link (rather than an AI-generated aggregation of sources) it may be the click through rate on these is higher, and will give good insight into how those searching are adapting to the rich snippets.
Read more
- X: Alex Greeland
- X: Rajan Patel
- XCancel: Whole Thread
ICYMI: Scary personal
Google recently launched Personal Intelligence (in beta) for AI Mode in search and results so far have been… interesting.
Personal Intelligence allows users to add their Gmail and Google Photos accounts to the search experience, and it’s powered by Gemini 3 for seamless searching and highly tailored results.
Early tests by The Verge revealed that it’s much improved when it comes to understanding personal interests, adding events to the Calendar, or updating a shopping list in Keep. However, it still struggles with the finer details, like sending people to closed shops or creating bike routes that seem less than safe.
the better web co. insight
If Google can make AI Mode feel truly personal, that will radically shift the way people search, as the personalisation will encourage more people to use this to get the unique answers they’re looking for.
But as results are supposed to be more personal, does that mean users will trust them more? That could be problematic as it still gets things wrong, such as this Frankenstein key lime pie recipe – the clear issue here is that the devil is in the detail. It’s like Siri in its early days (and still, to a degree now): after the third time of it not accurately delivering what you’re asking for, you’ll just decide it’s not for you and that’s the end of your use of the service for a few months (at least).
Read more