20+ years helping websites get found in search. 19+ years trying to explain it to my family. Director SEO at @Ticketmaster | @LiveNation Entertainment.

Los Angeles, CA
Joined May 2010
Robert Ramirez retweeted
Very cool: This study analyzed conversational queries from 150+ Search Console properties: It found: - The median number of AI queries grew 7x - Conversational replies (yes, yes please) grew 21x - Follow up questions (what about salary) grew 39x - Only 0.13% of total queries were AI
2
3
21
1,484
Robert Ramirez retweeted
My ideal AI SEO tool isn't a prompt tracker. Instead, it: 1) Tests exactly what AI gets right + wrong about a company's brand facts 2) Measures overall AI recommendation rates for each company/competitor 3) Importantly, crawls and analyzes all pages to recommend specific content changes So I built *exactly* that tool with @letaidocom as an #ahrefspartner, using data from Brand Radar to generate client action plans • Tracks what AI knows about the company, services, and brand facts • Measures how often AI mentions the company 📈 or competitors 📉 for every service • Analyzes WHY competitors get recommended • Finally, identifies specific information gaps, weak on-page evidence, and pages to target We can share this report with clients/stakeholders/content managers so they understand exactly what's really driving their AI answers - people *really* start to understand when you show them this. Previously, I could get much of this data in @ahrefs Brand Radar, but turning that data into specific fixes required more manual work I've REALLY enjoyed building these tools (Full prompt in thread so you can build this yourself)
18
9
1
82
6,037
Some rank trackers charge top $$$ and while this isn't a problem of their own doing, it sure does cast a shadow on the value they offer. Creating dashboards and rank tracking from GSC data isn't a value add. I can (and already) do that myself.
32
Robert Ramirez retweeted
Recommended reading if you want to understand the very likely future of Google's Generative Search Results @VeryWellVersed even included a tool to show you what your search results may look like in the future
New post: How Do We Move Forward if the SERP Is Generative? ipullrank.com/generative-ser… I’ve been thinking a lot about the poisoned results thing @Derek_Perkins posted about the other day. Google introduced something called “generative UI” a while ago. If this is the tech powering the SERPs we are in for a crazy ride. I built a proof of concept called “Genraggle” to help you get a sense of how crazy probabilistic SERPs might get.
3
8
75
7,261
Robert Ramirez retweeted
New post: How Do We Move Forward if the SERP Is Generative? ipullrank.com/generative-ser… I’ve been thinking a lot about the poisoned results thing @Derek_Perkins posted about the other day. Google introduced something called “generative UI” a while ago. If this is the tech powering the SERPs we are in for a crazy ride. I built a proof of concept called “Genraggle” to help you get a sense of how crazy probabilistic SERPs might get.
9
8
3
58
11,361
Robert Ramirez retweeted
It's worse now -> Quick update on the ChatGPT utm situation. Yesterday I was still seeing utm parameters for citations in the "More" section of the right sources panel, but not anywhere else (in the answer or core sources). But now it's worse... as of this morning, I'm not seeing utm parameters at all. If you check GA4, you will see a big drop in ChatGPT visits based on this change. Again, I think this is a bug... it doesn't help ChatGPT at all, and I hope we see this change back soon.
I'm still seeing utm parameters, but not for the main sources, only for the "More" section. Super weird. Maybe a bug? Time will tell.
6
2
2
20
4,564
Robert Ramirez retweeted
SEOs ❤️ Bing
INTERESTING! Bing Webmaster Tools is once again leading the charge: you can now see topics you're being cited for, how often they're cited and related grounding queries. This is the closest we have to classic search data. Your turn, Google!
17
17
1
189
11,139
Robert Ramirez retweeted
Google disabled Google search in my @GeminiApp because I added an attachment, it said: "Uploading a document to Gemini Chat now locks the session to those files to prevent confidential document content from leaking into external search queries and to stop outside web results from contaminating your file data." Somebody thought this was a great idea and it passed all the checks and went to production. WOOOOOOOOOOOOOOOOOW.
15
14
6
311
41,811
Robert Ramirez retweeted
Long live the link! Google is updating how it handles links in AIOs and AI Mode again. Barry has the details: *Google increased links in AI Mode and AI Overviews. *Google is placing some of those links higher up in the AI responses. *Google will show one or two links to the "right web page" for some queries that deserve it. *Google is expanding the timely article carousels to more regions and languages globally. *Google has more prominent link designs for related articles. *Google is testing linking Search profiles from creator and publisher names on videos within AI Mode, but those links go to Google pages, the Search profile page, first.
5
7
1
62
5,966
Robert Ramirez retweeted
Google doesn’t label AI Mode queries in Search Console, but they leave a trace: people reply to an AI, and the reply gets logged as a search: @jwbobbink classified every query of 155 sites over 16 months to see how often that happens, and how fast it’s growing 👇 Here are key insights: * Replies typed at an AI are now normal. Bare replies such as “yes”, “ja” or “yes please” showed up as searches on 94 of the 155 sites. Per million impressions they grew from 5.6 to 116: 21 times as many in 16 months. * German speakers say please to an AI, Indonesian speakers don’t. Three in four German yes-replies come with “bitte” or “gerne” (“ja bitte” on 52 sites). Dutch speakers add “graag” two times in three, English speakers say “yes please” 40% of the time, and Indonesian speakers never do. * Follow-up questions are rising from a small base. Queries like “what about salary” are still rare, but they grew 39x over the window, most of it after May 2026. * The typical site sees seven times more AI-shaped queries. The median site’s AI-shaped share of impressions went from 0.032% to 0.22%. Across all impressions pooled the share fell, because a few very large sites carry most of the volume. * AI-shaped queries rarely get clicked. Their click-through rate was 2.9%, against 35.6% for all queries pooled. * It’s still small. AI-shaped queries were 0.13% of all impressions. The growth is the story, not the size. More: gscwizard.com/learn/research…
28
14
2
73
6,508
Robert Ramirez retweeted
Excellent coverage by @nozzleio sharing how Google is poisoning the SERPs that it serves to rank trackers since last month ... so if you have seen rankings "off" in the latest weeks: this is your confirmation: Google has been poisoning the SERPs it serves rank trackers: complete, normal looking pages whose results have been swapped out. Nothing on the page gives it away. Each keyword’s own history does. From 27.5M analyzed SERPs: 8.0% of SERPs came back poisoned, the page looks normal. The results don’t, No error, no captcha, the same template, but most of the keyword’s usual sites are gone, the organic results keep 20.7% of the previous SERP’s sites, against 88.5% before. * Social sites and forums fill the gap: Reddit, Facebook, LinkedIn, Instagram and Quora show up more often; YouTube, Yelp and MapQuest less. * Tracked sites fall out of the top 10: A project’s own site misses the top 10 on 58.6% of poisoned SERPs, against 2.7% before. * It passes within days: 77.9% of poisoned SERPs have their usual sites back within three days. A one-day drop in late September is more likely this than a ranking change. Read more here: nozzle.io/research/poisoned-… PS: Thank you so very much to the Nozzle team to let us know 🙏
23
16
2
88
7,036
Robert Ramirez retweeted
Before you report an unconfirmed Google update or panic over a major drop in your rank tracker: check if you were served a poisoned SERP. An interesting new study from @nozzleio looked at 27.5M SERPs across 49 countries and found Google has started actively poisoning results served to scrapers. Instead of blocking requests with 429s or CAPTCHAs (which just tells scrapers to rotate IPs and retry), Google serves a fully rendered, completely normal-looking SERP. What this is resulting in is that the organic results are quietly swapped out with irrelevant data. Some key observations from the study: • 8% of all SERPs were poisoned post-Sept 13 (peaking at ~27% between Sept 27–29) • Organic results kept only ~21% of their usual sites (vs. 88% normal) • Tracked client sites fell out of the top 10 on 58.6% of poisoned SERPs • Social sites like Reddit (+8 pts), Facebook (+5.5 pts), and LinkedIn surged to fill the gaps • Interestingly, paid modules (Ads and Shopping) remained almost completely intact, while AI Overviews and PAA dropped significantly • Desktop pagination was heavily truncated (stopping at page 4 or earlier on deep queries) This is essentially Google's countermeasure against automated AI scrapers and RAG grounding pipelines, but rank trackers are inevitably being impacted. If your client's ranking report shows a significant decline overnight, check GSC average position first. If GSC hasn't displayed a major deviation, it is likely that it is a data anomaly rather than a change in your ranking Read the full write-up and data: nozzle.io/research/poisoned-…
31
40
2
191
15,650
Robert Ramirez retweeted
Incredible: Not only is Google working to block AI crawlers, but @nozzleio uncovered evidence that Google is deliberately serving FAKE search results to scrapers The "poisoned" search results look real, but w/ garbage data. This influences AI crawlers, scrapers, & rank trackers.
124
241
92
1,782
147,770
Robert Ramirez retweeted
Finally, @seroundtable published a story on this after @nozzleio published original research: seroundtable.com/third-party… This wrecks the data in tools like Ahrefs Site Explorer and SEMrush. It's now all suspect and must be heavily scrutinized before used to make biz decisions.
Something is definitely wrong with @ahrefs rank tracking today. Every site we work with has "tanked" in Ahrefs, but when we spot check, rankings are fine. Realize this is a challenging time for rank trackers in general, but would appreciate proactive comms from the Ahrefs team.
8
4
2
47
6,672
Robert Ramirez retweeted
New: Google launches UGC Fresh Data Program to send your fresh User-Generated Content directly to Google to h̶e̶l̶p̶ ̶p̶e̶o̶p̶l̶e̶ ̶e̶a̶s̶i̶l̶y̶ ̶f̶i̶n̶d̶ ̶y̶o̶u̶r̶ ̶c̶o̶n̶t̶e̶n̶t̶ gobble it up for AI training and click-stealing answers
23
21
3
159
15,422
Robert Ramirez retweeted
Yep, like we've been saying for a long time, the AI platforms are NOT going to let the gaming of AI Search continue to happen. BEWARE. Do Not Engage in Deceptive Campaigns or Artificial Activity: *Manipulate the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship, or independence (e.g., networks of sites posing as unaffiliated sources corroborating the same claims) *Build, improve, maintain, or contribute to tools, systems, or infrastructure designed to facilitate or enable deceptive campaigns, including astroturfing, sockpuppet account creation, coordinated inauthentic behavior, or scaled impersonation And more...
Using Claude to try to influence AI responses? This is a great catch by @HannsKronenberg and @MattGSouthern: @AnthropicAI updated its Usage Policy a few days ago (announced Oct 8, takes effect Nov 12) A new section bans using Claude to: ❌ Seed the sources that search engines and AI tools pull answers from with content that hides who wrote it or who's behind it (their example: networks of sites posing as independent sources backing the same claims) 👀👀 ❌ Push content through networks of fake or "independent" sites to hide that it comes from one source 👀👀 ❌ Build or sell tools that do any of this (!!) The previous version (Sept 2025) *didn't mention search engines or AI answers at all.* Fake reviews and fake personas were already banned. The penalty is losing access to Claude. Feels a lot like the early days of Google’s spam policies!
14
1
1
26
6,486
Robert Ramirez retweeted
Holy moly SEOs!! Google just gave us a goldmine of data on how long it takes to crawl URLs, index content + recover from core algorithm updates:
23
64
9
566
34,930
Robert Ramirez retweeted
A few folks who shared my recent article on "main content" posted it and used a screenshot of this table. Here's something cool. After I finished writing my article, I asked Antigravity what we could do to follow my own advice and make the main content - the reason why people were reading this piece - easier for people to ingest. Antigravity suggested and create a table. I made a few modifications on it and then put it at the top of my article. Here's a good prompt you can use in your LLM: What is the primary purpose of this page? What is it helping people understand or achieve? Give me three ideas on how we could improve or add something to the page that would make it even more effective at achieving its purpose and helping people.
9
18
1
95
5,083
Robert Ramirez retweeted
This is freaking sweet SEOs. Help Center + documentation is often a HUGE source of citations for LLMs. You can automate creating them by leveraging the Claude browser extension. Claude will agentically run your tools, document them + take screenshots.
4
5
1
81
5,961
Robert Ramirez retweeted
Whoa, GSC is now telling site owners how many people are adding them as 'Preferred sources'. Just received a flurry of emails from Google. This would be cool to see *IN GSC*. :)
29
34
12
251
30,331