November 19, 2025
The marketing industry has been hit with an endless stream of 'SEO is Dead' headlines in recent months. This has been prompted by fears that AI (specifically Large Language Models (LLMs) like ChatGPT, Gemini, and Grok) is rewriting the rulebook entirely, and that everything we know about organic search is set to change as more people use LLMs for search, and Google increasingly pushes AI overviews in the SERPs.
However, Rodrigo Stockebrand, former head of of SEO at huge names including NASA and Amazon, recently posted a breakdown on LinkedIn that caught our attention.
In his post, he looked at a specific page from Grammarly’s blog - a simple long-form article about the dozens of birthday wishes you could write in a card, which Ahrefs data shows this is one of the most cited sources on the web by LLMs.
But why is this one of the most cited sources?
Stockebrand deconstructs why the bots love this page so much. And although he used terms like 'lexical simplicity' and 'ontological breadth' which may be new to some SEOs, the principles aren't new. What makes an LLM want to reference your content overlaps significantly with the things Google has been asking for, for years.
So, let's dive into his findings and what they mean for you.
Stockebrand noted that the Grammarly page is written at a Grade 6–8 reading level (early secondary/high school level). He explains that, from a technical perspective, this helps AI because simple sentences are easier to 'tokenize', which means it’s less complicated for the machine to break down the data without getting confused by messy syntax or idioms.
Although in some niches it’s important to showcase your knowledge through highly technical writing, readability has always been a crucial element of content writing. User engagement matters for SEO, and a wall of dense text is usually not the best way of capturing attention. And as suggested in these findings, it's now the best way to keep LLMs happy, too. If your writing is overly complex for the topic you’re covering, it can be harder for the AI to process your point.
Another point the analysis covers was 'clear term distribution.' The Grammarly page doesn't just mention 'birthday' once, it creates a web of related terms like funny, heartfelt, romantic, and inspirational, that all link back to the main topic. To an LLM, this creates a strong pattern and proves that the page is definitely about birthday wishes.
Of course, this is exactly how Google has kept keyword stuffing dead and buried, by expecting on-theme content and related phrases. All good SEOs know that it's crucial you include semantically linked keywords while showing Google the content is natural.
Additionally, 'redundancy for reinforcement' and 'term-frequency' help LLMs be sure about the content they’re referencing, which sounds a lot like ensuring important keywords appear regularly throughout content - a standard SEO practice.
Perhaps the biggest reason the Grammarly page succeeds is what Stockebrand calls 'ontological breadth'. Essentially, the article is comprehensive - it covers birthday wishes for your mum, your boss, your neighbour, and even your ex.
Since LLMs function as answer engines, they want to source data from the most complete 'node' available. If your page only covers half the topic, you aren't the best source.
This is again familiar. Skyscraper content (creating a superior resource that covers every basis of a topic) has been a staple of content marketing for a long time. This analysis simply validates that strategy. And of course, content that covers more will hit a higher number of keywords, ultimately resulting in more clicks.
The shift we're seeing isn't necessarily in how we write, but in why we are being read and who it is for. In the past, we wrote great content for people while keeping Google in mind. But now, we are being mindful that content is also used as training data for LLMs.
The Grammarly page isn't just a list for a human to read, it’s now also a structured database that an AI can learn from. When ChatGPT needs to write a funny birthday card message for a coworker, it references the Grammarly page because that page has perfectly categorised what 'funny' and 'coworker' look like in this context.
The takeaway for us isn't that we need to do things differently. It’s that there’s now even more reason to get content right and avoid cutting corners. The fluff that used to pass for content is going to get ignored even faster by LLMs than it was by Google.
If you’ve been doing SEO the right way, by focusing on real quality rather than corner-cutting, then chances are you’re already optimised for AI.
If you're looking for expert help navigating the world of LLMs and AI, don't hesitate to get in touch with us today.
