Terry Van Horne AKA Webmaster T
@terryvanhorne.bsky.social
Latest project AIgroundingTools.com measuring AI visibility by the byte using Gemini Grounding API! Proud Leaf fan & SeoPros Founder & Co-Founder of SEO Training Dojo living one day at a time doing what I love in the greatest country in the world!
From what I've read and my feelings about the ethics of these content thieving bastards in the near future they'll be a pay (advertise) or go away when it comes to citations and mentions in their results. You know what I'm saying! Start making a case to ignore them & block them when that time comes!
I follow Wil closely especially for his work on developing new KPIs that IMO are important to the industry going forward. This week I'm going to begin developing a tool close to what Wil did by using LLMs to generate fan-outs & prompts Fan-out research is to AIO what keyword research is to SEO.
Wil Reynolds@wilreynolds.bsky.social · 2w agoI ran a bunch of prompts to see how fan outs search vs how I write my content. Learned that there's a huge gap between the two. Is your "great content" loved by humans but is not visible in ChatGPT? What would it look like to optimize content taking into account fan outs. youtu.be/xACZt4Xr8ec
For the past 6 months I've spent a lot of my time working w/Gemini Grounding API. The biggest eye opener for me was how many contributions to an AI answer are never cited little on mentioned which means the current method of measuring visibility via scraping results in ILL CONCEIVED KPIs!
This is why I believe prompts are ineffective for tracking visibility and why I don't spend time crafting them. IMO, the 15% of queries cited as never seen before by Google could be the entire extent of the usage by users. Current AI KPIs are ill conceived by trying to put square pegs in round holes
Dan Petrovic@dejanseo.bsky.social · 2w agoReading and processing. aimclear.com/prompts-for-...
There have been "rumors" they were developing their own search index. I'm leaning towards this was actually what they were up to.
Glenn Gabe@glenngabe.bsky.social · 3w agoScary -> OpenAI says it will no longer train its AI models on the live internet, citing “significant risks”. "OpenAI said that it will train its models using cached copies & simulated versions of the internet instead of the live web b/c the models are too unpredictable to be trusted."
This has been on my mind since I read it and combined w/what I've read from others (I highly respect) like Mr. Tabke claiming that organic results will either be moved to a tab or just disappear seem to have underestimated the users role in these decisions. www.seroundtable.com/ai-search-st...
New Studies Say AI Visibility & Search Not Worth Investing In
There are three new studies out over the past week that suggest that spending time, money and resources on AI visibility and AI search may not be worth it. In short, the studies all show that AI Sear...
seroundtable.com
Before you go turning all your sites off you may want to give this a read WITH AN OPEN MIND! The author makes a good point about the amplified hysteria caused by using LLMs as the ONLY source of info. IME, LLMs may be trained, out of caution, to answer a certain way on topics like this
Joe Hall ☕@joehall22.bsky.social · 3w agoAn over due reality check. eliastaylor.com/noexploit/
Do read this! I especially like how Ryan took info from Perplexity and showed how it applies to all AI. Like SEO since hummingbird it is more important to get entity usage for the corpus not a single page.
Ryan Jones@ryanjones.bsky.social · 4w agoDon't sleep on this post. It's a proven process to get cited by LLMs.
IMO, Bravery or "thinking outside the box" is what sets all marketers apart from one another. My most effective campaigns were often based on those 2 things. Convincing a client to go all in on video day Google bought YouTube & before video hit Google SERPs. When they hit SERPs we had over 100 @ #1
Jono Alderson@jono.id · last mo.Bravery should be marketing's most important KPI. Unfortunately, it doesn't have an ROI forecast, so the board won't approve it. We reward fictional certainty, punish honest expertise, and call ourselves 'data-driven'. www.jonoalderson.com/conjecture/b...
Trump Tariff Math: "Canada’s trade surplus with the United States grew to $11.2 billion in August from $6.1 billion in July" Imports to Ca. from the US were down 2.1% ~ yeup because we can do w/o US consumer goods & US cars! Trump says US doesn't need Ca. products? Seems it's Ca. doesn't need US
It's not just AI related. Poorly done programmatic sites would fall into this category as well. AI Effort != work is just the easiest one to identify.
Cyrus Maxx@zyppy.com · last mo.Helpful Tip: Since Google updated its docs on Helpful Content, lots of folks are talking about "Effort", along w/ fake author profiles Effort ≠ work In the age of low-effort, scaled AI content getting decimated, understanding “high-effort” is increasingly relevant signal.zyppy.com/p/content-ef...
Sometimes it's back to the future in the SEO game! In the 90's SEOs built a page for each of the 4 or 5 engines. I never did this because I figured out there are overlaps in what to optimize w/the difference being how the elements were weighted for ranking. I think that holds true for AI as well!
Some great points here one I would add is part of the reason for SEOs becoming all things web related is clients see them as their "web guy". Good SEOs help those clients find the right person for the job.
Joe Hall ☕@joehall22.bsky.social · last mo.ashla.ai/p/opinion-se...
Nice to see someone else who questions prompt tracking believing that tracking the core entities yields better info on what and how mentions get into AI answers. Entities are just as important for AI (maybe more) as search.
Dan Petrovic@dejanseo.bsky.social · last mo.I meant to write a lot more, but this will do for now: dejan.ai/blog/prompt-...
This is so true especially given Gemini 3.8 (the brains behind Google AI Mode) has increased use of multi-modal content. I have been comparing 3.5 Gemini to 3.8 and in a very limited test YouTube DOMAIN often has by far the highest share of voice.
Majestic@majestic.com · last mo.Katherine Nwanorue believes that SEOs and marketers have always prioritised multimodal content by repurposing it across several mediums, starting with a blog post and then turning it into videos and infographics. maj.to/4abzPj2
Wil recognizes there is change needed for AI KPIs. Bottom line is just because you weren't mentioned doesn't mean your content wasn't used by ChatGPT to synthesize the answer! So mentions in scrapped data doesn't cut it! SEO KPI metrics reflected the quality of work. AI KPIs are vanity metrics!
Wil Reynolds@wilreynolds.bsky.social · last mo.Here are 5 KPIs that will help you convince the c-level that visibility, citations, etc are the wrong metrics, and these instead will lead to sustainable, long term visibility. www.seerinteractive.com/insights/ai-...
Users are no longer sold by informing them about a product or service because an AI result has already provided that when they get to your site. Duane outlines how to decide what users now need from you when they arrive at your site regardless of how they got there
Duane Forrester@duaneforrester.bsky.social · 2mo agoAI changed the buyer. They can now research products, compare prices, check reviews, identify risks and narrow a shortlist in minutes. That changes what they need from you when they arrive. Read: duaneforresterdecodes.substack.com/p/ai-changed...
Looking at the Gemini Grounding API you quickly understand "AI Contribution Pilot" Publishers have an audit trail for contributions in the Gemini Grounding API. Google saved publishers once with AdSense which monetized content when newsletters and begging just weren't cutting it. History Repeats!
I strongly suggest reading this post it outlines user behaviour that explains why current AI Success KPIs were ill conceived in particular visibility metrics of citations & mentions. Bottom line is users don't use them & CONTENT EXPECTATIONS when they do arrive at sites have changed significantly!
Duane Forrester@duaneforrester.bsky.social · 2mo agoLLMs are time machines. They take present-you, you with a question, and hand back future-you, the version with a decision. What gets lost is the research journey, including the signals that told you how much confidence the answer deserved. Read: duaneforresterdecodes.substack.com/p/llms-are-t...
SEO tool providers must be having real problems getting Google SERP links true destinations. I ran 400 yesterday and 96 returned 403 or 429 response codes. Fair enough! I was doing head checks to lessen the load for both ends so I ran another script asking for a partial. 6 came back 206 and 4 404s
I have been pretty much of the opinion that claims of "fair use" were nonsense and the content thieves knew it but were/are counting on expensive litigation to lessen their risk!
Samuel Lavoie@samuellavoie.bsky.social · 2mo agoAn internal Microsoft document said generative AI products have created a “doom loop” that is killing “the entire web.” No wonder AI companies wanted to hide this from the public. These indictments point to how LLMs were trained, how they work, and the threat they pose to human labour.
No click SERPs have been the Google goal since 2006 w/purchase of Orion patents began a transition from a pure lexical indexer to an entity based answer engine with an understanding of concepts and relationships. That we are where we are shouldn't be a surprise! It's only going to get worse!
Spent some time scrubbing Google goto links from grounding data for study of: 1. How well sites in Google AI results rank for target terms & grounding queries (fan-out) 2. Compare use of parametric memory of 3.5 to 3.8 3. Determine source overlap of 3.5 to 3.8 4. Sensitivity to triggers 3.5 to 3.8
Joe makes great points on many of the "GEO bros" techniques. Many are clearly spam that work only because the AI providers aren't yet equipped to handle it. They will because in the end the public starts talking about their shite results and survival mode kicks in! ashla.ai/p/is-your-ge...
Is Your "GEO" Just Another Name For Spam?
You might be making yourself a target for the next spam update!
ashla.ai
Interesting post by Pedro Dias about measuring AI visibility. I've been comparing output from Gemini 3.5 and 3.8 which indicates the problems Pedro is discussing will not improve they're quite likely becoming the "norm". theinference.io/p/it-was-the...
It Was There a Minute Ago
A missing brand mention gives you a result. The explanation attached to it deserves a much harder look.
theinference.io
It was obvious to me that stealing content and bandwidth w/little to no benefit to the site wasn't a sustainable business model for a search platform. It appears OpenAI/ChatGPT have been building an index since 2024. Whether that decreases site fan-outs is TBD in due time peec.ai/blog/chatgpt...
ChatGPT built its own search index - Peec AI
ChatGPT uses Google and Microsoft, but owns its own index Read more on the Peec AI blog.
peec.ai
The post where I replied saying there was no change in grounding data wasn't accurate I'm currently in the midst of preparing to compare output from 3.5-flash and 3.8-flash.