Pinned Tweet
Must have SEO extension with AI assistant chromewebstore.google.com/de…
Serhii Koksharov 🇺🇦 retweeted
I'm 51.
Solo founder from the Netherlands.
Building privacy-first web analytics in the EU.
Looking to connect with founders and indie hackers who actually look at their numbers.
Serhii Koksharov 🇺🇦 retweeted
Lastly, with all these updates, we are going to be sunsetting Atlas.
All these capabilities were built on what we learned from Atlas users who took a leap of faith on a new browser.
You taught us how agents can help make browsing and doing work on the open web better, and we are applying these learnings to these new products.
The current targeted date for deprecation is 8/9, and we'll share more information in the upcoming days both in-app and via email.
Thank you all for your support and we hope you love these new features!
Serhii Koksharov 🇺🇦 retweeted
Cool post from @dejanseo -> How AI Search Grounding Actually Works: Google vs OpenAI vs Anthropic
"For Google, the funnel barely narrows: 7 received, 7 cited. This is the defining trait of Google’s grounding — it doesn’t expose pages it didn’t use."
"OpenAI’s wide-receive / narrow-cite split is a visibility trap. Being one of the 37 pages OpenAI read is very different from being one of the 2 it cited. Both are worth tracking, but they’re not the same win."
"Anthropic gives the richest “considered set” (cited + rejected) but hides the snippet text, so reconstructing what it actually read costs a second pass — and real time and tokens."
dejan.ai/blog/grounding/
My theory on why Google quietly killed the inurl: operator.
Vibe coders ship fast but don't think about security. Their apps work — but rules/*.md, .env, config.ini files are wide open.
API keys. DB passwords. Just sitting there.
And inurl: was a neat way to find them.
Combine that with SEOs using it for grey-hat tactics, and Google had two good reasons to pull the plug.
At some point I needed to collect event data and keep it updated regularly. So I decided to build an agent that does it for me:
- Scrapes event listings and parses the content
- Sends the aggregated content to an LLM for deduplication
- Searches the web for the events' official websites
- Visits each site to extract details
- Collects logos separately
- Then structures and stores everything
Sounds cool, but there are a lot of caveats. This is where I truly understood why search engines need structured markup — and why search algorithms aren't perfect.
The main problems:
1. Websites block access or serve content via JS.
You need pretty advanced crawlers to access all the content across different sites. That gets expensive resource-wise.
2. Event names sometimes match the names of completely different entities.
As a result, the LLM picks the wrong site as the "official" one. To make it work correctly, you have to search for the event using 2–3 different queries, visit several sites, and evaluate each one against a set of criteria to confirm it's actually what you're looking for.
3. Some events don't have a website at all — tickets are sold through aggregators or listing platforms.
4. Organizers rarely structure the information on their sites.
Even humans struggle to figure out the exact dates, whether it's a casual meetup or a full conference, who the organizer is, and so on. LLMs get confused too and often hallucinate due to the lack of clear information.
5. A missing price gets interpreted as a free event.
In reality, the organizers have just announced the event and haven't published pricing yet. So paid events end up classified as free.
6. Many events have no logo — just a cover image or banner.
The LLM goes looking for og:image as a reliable source of the event's visual identity. But what's there is often not what you'd expect, or the OG tags don't exist at all. You also need image recognition and have to go through several candidates to find something relevant to the event. A human would do this faster.
And there's a whole pile of other quirks that search algorithms have been dealing with for ages. Meanwhile, all it would take is to:
- Make your content accessible
- Structure your text content and avoid ambiguity
- Use data markup — Schema and OpenGraph
- Pick a proper, unique brand name instead of a generic one
- Make sure all of this actually keeps working
Hopefully this post is useful not only to event organizers.
Serhii Koksharov 🇺🇦 retweeted
Chrome 149 DevTools is out → goo.gle/4v55cVT
Here's what's new:
💡 Agent walkthrough & Copy to coding agent in AI assistance
🎨 CSS Code Completion from Gemini in the Styles tab
🤖 WebMCP debugging tools in Application panel
🚀 DevTools for agents 1.0 with custom tools
It looks like Google has deprecated the inurl: operator. It no longer works for me. @rustybrick is it new?
Serhii Koksharov 🇺🇦 retweeted
SEO Tip: You can use Ahrefs "Crawled Pages" report to quickly identify your competitor's current content strategy.
You can see all the pages that were most recently indexed by the Ahrefs crawler.
Serhii Koksharov 🇺🇦 retweeted
Google anuncia, en colaboración con la comunidad de Schema(.)org un nuevo repositorio con datos de uso de los diferentes marcados de datos estructurados a lo largo de Internet.
Estos datos, que se actualizarán de manera mensual, estarán disponibles para su análisis o descarga (en .csv o .json) desde el GitHub oficial de Schema: github.com/schemaorg/schemao…
Una manera de ver la adopción de los datos estructurados con el paso del tiempo, así como para ver cuál de ellos escoger en base a su nivel de implantación en casos de duda ante otros similares.
Puedes ver el comunicado oficial de Schema en: blog.schema.org/2026/06/04/a…
Así como algo más de información adicional en: schema.org/docs/usage_stats.…
Nice catch from Lily Ray!
The future of SEO isn't just about keywords anymore - it's about building trust through authentic influencer partnerships and real human expertise.
Called it 😅
youtu.be/KFY-84CqejA
Serhii Koksharov 🇺🇦 retweeted
JUST IN 🚨 Google to show AI Search performance data in Search Console #SEO
"To help you understand how pages from your site are shown, our new reports show the following information:
Impressions: How often URLs from your site appeared in generative AI features in Search and Discover.
Pages: Check which URLs appeared within AI features.
Countries: Understand your visibility on a country basis.
Devices: Identify the devices people are using when seeing your website (available for Search results).
Dates: Monitor your performance over time with hourly, daily, weekly, and monthly granularity."
Serhii Koksharov 🇺🇦 retweeted
Someone on Reddit asked 4 LLMs 100 times each: walk or drive to a car wash 100m away?
Gemini 3.1 Pro said drive every time. Claude Opus 4.8 said walk every time.


























