Latest Organic Search News – August 26

We’ve compiled the essential updates from Google and the broader world of search from the last month – keeping you up to date with everything you need to know.

TL;DR

• Search Console now reports on your social and video channels. Platform properties went globally live on 29 July. Claim your YouTube, Instagram, TikTok and X accounts and you can see how that content performs in Google Search.
• Visitors from AI tools become leads 3x more often than other organic traffic, but they’re only 0.5% of sessions. Orbit Media analysed 97 B2B sites and 29 million visits. ChatGPT sent 82% of that AI traffic.
• Yelp is licensing its reviews and photos to OpenAI. ChatGPT’s weakest area has been local queries, and it has just bought its way out of that. Quote requests are coming too.
• AI cites fresh pages, but the freshness comes from updating old ones rather than publishing new ones. In Seer’s study, 75% of cited pages had been updated in the past year while only 42% had been published in it.
• Most product categories still have no brand that ChatGPT reliably recommends. Across 1,094 categories, only 15.2% had a clear leader. Once a brand takes one, it tends to keep it.
• Publishing your own data earns a page 3.3x more AI citations than average, but one format does nearly all of it: the benchmark study that ranks named competitors against each other on something measurable.
• Google is moving the Top Stories news carousel up into the AI Overview itself. That happens on 17.5% of UK news searches which show the carousel at all, and publishers who opt out of AI features look set to lose the slot.
• Google switched on AI Overviews and AI Mode in France on 22 July, two months early, promising publishers an opt-out, separate reporting for AI search, and continued compensation.
• Google Search traffic to publishers fell 34% over the past year, and the smaller the publisher, the harder it landed.
• AI agents still can’t reliably read pricing pages. When they fail to extract a price, 58% of the answer comes from third parties instead of you.
• Half a million sites block the crawler behind most AI training data, and most never chose to. The block usually came from a CDN or platform default rather than a decision anyone made.
• And a closing thought from Wil Reynolds: SEO let you rank for things your company wasn’t actually good at. AI search checks your claim against what everyone else says about you. Strategies must evolve.

Search Console can now show you how your social and video content performs in Google

Potentially the most useful launch of the month.

Google’s platform properties went globally available on 29 July. If you manage a YouTube, Instagram, TikTok or X account, you can claim it as a property in Search Console and see how that content performs within Google Search: impressions, clicks, queries, countries, devices. You add it the same way you’d add any new property.

Google has also published a help document on what to do with the data. The use cases worth knowing about:

• The Insights report shows which query groups are sending Google traffic to your social and video content, so you can see what your audience is searching for before they land on a video.
• A 24-hour filter flags sudden spikes on recent posts, which is a useful cue to cross-promote something that’s taking off.
• Multi-property export lets you pull YouTube, Instagram and TikTok into one spreadsheet and compare them side by side.
• Search Console annotations work here too, so you can track whether rewriting a YouTube title or caption actually moved anything.

Until now, the way to answer “are our videos showing up?” was to scrape SERPs and check whether video results appeared and which ones ranked. That gives you good data, but it only covers the keywords you thought to track. You’re inferring a picture from a sample you chose in advance.

Pulling this into Search Console changes the shape of that:

• You get the whole query landscape, not just your own keyword list.
• You can see how often your videos and social posts actually appear, rather than estimating it.
• You get real click and impression data behind those appearances.

That’s a considerable step up, and it ties directly into the Search Everywhere Optimisation approach we take at Distinctly: your audience isn’t only searching on Google so by being strategic, you can pick up both views and customers from other channels plus take up more SERP-space at the same time.

There’s no backfill, so history starts the day you verify. Connect the accounts now, even if you’re not ready to act on the numbers yet.

AI referral traffic is still tiny, and it converts three times better

Orbit Media’s new research, published 22 July, is the latest dataset on a question plenty of people have been eyeballing in their own GA4.

Orbit analysed 97 B2B lead generation websites and 28.9 million sessions over the year to June 2026, manually categorising every key event as high-intent, low-intent or excluded. 

Visitors from AI sources were three times more likely to convert into a lead than other organic traffic. Take the per-site median instead of pooling sessions and the gap widens to 7x. The pattern held on roughly two thirds of sites individually, not only in aggregate.

Then the number that keeps everyone honest. Of 28.9 million sessions, about 140,000 came from AI sources. That’s 0.5% of traffic, one visit in every 200.

Two other findings worth pulling out:

• ChatGPT accounted for 82.3% of AI sessions and turned 2.1% of its visitors into leads, against 0.5% for Google organic. Google still sends roughly 100x the traffic, so it remains the far bigger lead source in absolute terms.
• Case study pages had the highest conversion rate from AI traffic, while blogs pulled the most volume.

Their explanation is the one we’d back. The consideration stage now happens inside the model: by the time someone clicks a citation they’ve asked for recommendations, compared options and narrowed the list, so they arrive for confirmation. AI is also filtering out people who were never a fit, which is part of why volume is so low.

This aligns with what we’re seeing across our own client base. AI sessions stay small enough that the channel looks marginal in a monthly traffic report, and the quality of those sessions is consistently better than their share of traffic would suggest. It’s a pattern we’ve been describing to clients for a while, and independent research at this scale is a useful thing to be able to point at.

The caveat of course is how murky attribution data is right now. Most users will research in an LLM, then head back to Google and socials to make further decisions before they convert.

So the role AI recommendations play in search is huge, yet the only numbers we have (referral) will always make it look like it’s playing no role at all. Please do not underinvest in GEO for these reasons.

It’s the same conclusion the Similarweb research we covered last month pointed to: session volume is the wrong yardstick for this channel.

Yelp is now feeding ChatGPT, and local SEO needs a wider brief

Local has been ChatGPT’s most obvious weak spot. It has now licensed its way out of it.

Yelp is licensing its reviews, ratings, photos and business information to OpenAI, giving ChatGPT real-time local recommendation data. Yelp branding and links appear alongside its content, though OpenAI controls the presentation. The deal doesn’t stop Yelp doing the same with other AI companies, and it already licenses to Yahoo and Apple Maps.

The part that should get local marketers’ attention is what comes next. Yelp’s Request a Quote feature is coming to ChatGPT local services searches, so users will be able to contact providers for quotes and appointments without leaving the chat. That’s a lead generation journey happening entirely inside someone else’s interface, on data you don’t own.

Your Google Business Profile has been the centre of gravity for local visibility for a decade. It isn’t losing that, but it is losing its monopoly. AI assistants are answering local queries from licensed third-party data, which puts the platforms holding that data between you and the customer.

The deal is US-first, but this isn’t American news. Yelp has been a standard citation source in UK local SEO for years and sits in almost every local audit we run, and local search features have a long history of landing in the US first and arriving here once they work.

The aggregators that matter in your market are part of your AI visibility. In the UK that means:

• Google Business Profile
• Yelp
• Trustpilot
• Tripadvisor
• The credible vertical directories in your sector

All audited properly, rather than set up once and left alone.

The old page you refreshed beats the new one you wrote

Seer Interactive’s study on content recency, published 24 July, is one of the more practically useful pieces of research this year. Sonny Vasquez pulled every page cited in non-branded LLM answers for four brands across four sectors, in ChatGPT, Gemini and Perplexity over four months. That gave 7,683 dated pages carrying 47,097 citations.

The headline: 75% of pages cited by LLMs were updated in the last year, 88% within two. Anything older than three years barely registers.

That much you’d expect. The useful finding is what happens when you separate last-update date from original publish date. Measured by last update, 72% of cited pages look fresh. Measured by publish date, only 42% do. More than a quarter of the “fresh” pages were first published over two years ago.

The freshness LLMs reward is being manufactured by updates rather than by new publishing. 

A second finding cuts against the instinct that newer is always better:

• Pages cited in only one of the four months were the freshest of the lot, at 86% updated in the last year.
• Pages cited in all four months, which are the ones you actually want, were the oldest, at 68% fresh with a median update age around six months.

A very recent update earns you the spike. Established pages kept reasonably current earn the staying power, and that’s a maintenance job.

One more thing for planning conversations: the content types held to the highest freshness bar, marketplaces and comparison pages, are mostly earned rather than owned, so you can’t refresh your way onto them. The practical steer is to stop putting a blanket refresh schedule on everything, look at which URLs are getting cited month after month, and let that set your threshold.

This is a shift we’ve already made in how we build content plans. Regularly updating and refreshing key commercial pages is now a standing part of our content offering rather than something we do reactively when a page starts slipping.

For a long time the default brief in this industry was a publishing calendar, with maintenance treated as tidying up. The data has moved the other way: the pages doing the heaviest lifting in AI answers are established ones that have been kept current, so the retainer work protecting them is worth as much as the next new article.

In practice that means auditing the commercial set first, agreeing which pages are genuinely load-bearing, and building the refresh cycle around those rather than spreading effort evenly across a blog archive.

Most categories have no leader in ChatGPT yet, and benchmarks are how you take one

Two pieces of research from Kevin Indig this month, and they’re stronger read together.

The first, with Semrush, tracked 1,094 US product and service categories in ChatGPT monthly from January to June 2026, covering over 50,000 brands and 600,000 citations.

A category here is a buying topic: the kind of thing someone prompts about when they ask which project management tool or which running shoe to get. A category has an “owner” when one brand takes the largest share of mentions across those prompts and sits clearly ahead of the runner-up.

In June, only 15.2% of categories had an owner on that definition. A further 53.7% were open fields with several contenders and nobody in front, and the biggest categories by AI search volume were among those without an owner.

Two supporting findings sharpen it:

• Ownership is durable once established. Brands with a five-percentage-point lead held first place in 90% of month-on-month comparisons, while narrow leads changed hands frequently.
• Domain-level SEO metrics don’t predict topic ownership. Authority Score and organic traffic correlated with it only about half the time.

That second point is the practical one. A brand can look strong on every domain-level metric in your reporting and own nothing in ChatGPT. The reverse also holds, which is the argument we made in June off the back of Away Resorts’ category leaderboard position: a challenger can take a category it has no business taking on brand awareness alone, provided it moves before the incumbents do.

Compare that with classic Google search, where the top of most commercial categories has been settled for years. Here, more than half the board is still open, and our honest read is nimble beats established more often than not.

So what wins a category? That’s where Indig’s second piece comes in. His citation analysis found that pages containing primary research averaged 11.3 citations each, against 3.4 for everything else. They made up only 2.7% of the pages in the set but earned 8.4% of all citations.

The qualifier is the interesting part. Those citations weren’t evenly distributed: the large majority came from benchmark studies ranking named vendors against each other on measurable specifications like speed, cost and latency. Strip that cluster out and first-party research barely registers.

So the bar is higher than “publish original data”. A survey producing an interesting statistic about your sector is a PR asset. A study that measures named things against each other on a specific yardstick and publishes the results is an AI visibility asset.

Publisher surfaces: Top Stories moves up, and France goes live

Two developments for anyone doing digital PR.

Axios published Chartbeat data on 31 July showing Google Search traffic to publishers has fallen 34% over the past year, and the distribution is regressive: across two years, small publishers lost 60% of their search referrals, medium 47%, large 22%.

Top Stories inside AI Overviews now has numbers behind it. Last month we flagged the test; there’s now data on how far it has spread. Press Gazette reported analysis from news SEO firm Newzdash, covering more than 17 million Top Stories instances across the 30 days to 28 July:

• Where a Top Stories carousel appears at all on a trending news search, it now sits inside the AI Overview 15.5% of the time globally and 17.5% in the UK. 
• Only the UK and US show the feature so far, out of 40-plus countries tracked.
• It replaces the standalone box rather than sitting alongside it, so publisher links move up the page instead of being duplicated.
• Entertainment is the category most likely to trigger it.

There’s a trade-off attached. Because the carousel appears either inside the AI Overview or standalone and never both, publishers using Search Console’s new generative AI exclusion will likely lose eligibility for the embedded placement. That control arrived in the UK because of the CMA, and it now has a visible cost.

One clarification from Newzdash’s John Shehata while we’re here, because it trips people up constantly. Blocking Google-Extended alone should not affect visibility in AI Overviews or Top Stories. Google-Extended governs model training and grounding; the Search Console exclusion is the separate and far more consequential control.

France finally got AI Overviews and AI Mode on 22 July, roughly two months ahead of the 23 September deadline Google had given publishers. France was the last major Western European market without either, held back by a neighbouring-rights regime stricter than the European baseline and a regulator that fined Google €500m in 2021 and €250m in 2024 over publisher compensation.

Google launched with three commitments to roughly 450 publishers:

• An opt-out that doesn’t affect classic search rankings.
• Separate reporting for classic versus AI search.
• Continued neighbouring-rights compensation for participating publishers.

Not everyone is satisfied. Press union SEPM says it only learned the exact go-live date on the day itself.

Which publishers opt in versus out is a live media-list consideration, not a background detail.

Two technical checks worth running

Two findings this month, both about how machines read your site.

Pricing pages, again. Last month we flagged Suganthan’s observation that ChatGPT failed to read a SaaS pricing page because the numbers loaded via JavaScript, and cited G2 instead. There’s now a dataset behind it.

Kevin Indig and Siteline’s David Kaufman ran agents against 100 B2B products with three buyer tasks each, repeated five times. Most sites were broadly agent-ready, with one clear breaking point: pricing. Three things caused it:

• Opacity. No price published at all.
• Poor machine-readability. Prices sitting in JavaScript, images or PDFs.
• Access friction. Fetch failures and bot blocks.

Only around 65% of plans had prices an agent could read directly. When the agent hit an access error, 58% of its answer came from third-party sources, presented as fact, against 12% on error-free runs. Publishing prices helps but doesn’t fix it on its own: plenty of failures happened on pages where the price sat in plain view for a human.

Common Crawl, and who’s blocking it by accident. Suganthan has also published a free checker worth adding to any technical AI visibility audit.

Common Crawl fetches over two billion pages a month and gives the archive away. It’s the raw material for most large training sets, and when Mozilla audited LLMs released between 2019 and 2023, 64% had trained on it in some form. HTTP Archive numbers show roughly 492,000 sites name its crawler CCBot in robots.txt, about 95% of them to block it.

The question is how many meant to. Since July 2025 Cloudflare has blocked AI crawlers by default on every new domain it serves, Squarespace ships its own default, and plugins add blocklists. His free tool reads Common Crawl’s archives to work out when a block first appeared, identify whose template wrote it, and probe for firewall challenges that robots.txt never shows.

Being in Common Crawl doesn’t guarantee you end up in a model’s training data, and being absent doesn’t guarantee you don’t. What it does give you over time is visibility: if your key pages aren’t being crawled, it’s probably not fantastic, but you’ll be able to see it and go and find out why.

The receipts problem: AI is harder to fool than Google was

A good closing thought from Wil Reynolds at Seer, and a counterweight to a roundup full of tactics.

His argument is that an honesty reckoning is coming. In classic SEO you could rank for things your company wasn’t much good at. With enough domain authority you could muscle to the top of Google carrying no evidence for the claim.

AI makes that harder, because it isn’t only reading your site. It’s reading Yelp, Trustpilot, Reddit and everyone else with a view on you, and it notices when your self-description and the wider consensus don’t line up.

Our own way of putting this: you could always polish a turd with traditional SEO. It’s much, much harder to do it in AI search.

Traditional SEO let you win on the strength of your own asset (your domain, your content, your links) largely regardless of what the rest of the web thought of you. AI search assembles its answer from everyone else’s opinion of you as well as your own, so the gap between what you claim and what the market says is now a visibility problem rather than just a brand one.

His test is a good one: “everything we say on our site needs a receipt”. Claim to be a leading voice in your sector, and the question becomes how many of the other recognised leaders would name you. If none would, the answer isn’t a content brief. It’s doing the work to earn it.

He also argues that ‘targeting a 30% citation lift by November’ (as a KPI) aligns a team on the wrong problem, and the better question is what you want to be known for being great at in five years.

Enjoyed this roundup? Stay updated with the latest organic trends, tips, and exclusive insights by subscribing to our newsletter.