According to my server logs they're all over the place hitting our robots.txt - and guess what? they've been around for so long, training ML models and LLMs.

Hold on a sec, where does llms.txt sit within all of that ahah

I was gonna catch up with the Godfather saga, but time is ticking ahaha

You're right, it was someone desperately trying to upsell a plug-in from their CMS.

Test built on stiff methodologies from another more traditional SEO test

will removing meta description lead to traffic decrease?

seodepths.com/seo-research...

"The schema markup (itemprop, itemscope, etc.) was there to label the data, but didn’t need it to understand or summarise the profile."

IMO, schema markup as data labelling is the best definition ever --- I'd borrow it from Jarno Van Driel from a recent podcast.

From the conversation, ChatGPT 5 went on:

"I relied on the plain visible content in the HTML page — the text inside headings, paragraphs, and lists."

I tested Sarah Chen

"""
✅ I used the human-readable content: “Sarah Chen — Chief Executive Officer & Co-Founder”

🔎 The schema markup was a signal about structure (helpful for machines, search engines, or AI testing) but not required for me to extract the details.
"""
chatgpt.com/share/68ac1c...

/3

It's probably our best bet to start optimising information retrieval with #AI engines

Lots of testing in sight though - AI mode vs Grounding Gemini API are based on different models, but to what extent is all to be found

/2
Where does this feed into your keyword research SOP for AI Mode?

Once you're familiar with the list of terms likely up for grounding (web search), customise the #Colab notebook and run them to get the potential fan-out suggestions

/1
In plain English,

Grounding is the process of turning long prompts into traditional search queries to pass on to Gemini's models to operate live searches in Google.