Exactly.
"If you’ve seen your clicks from Search dropping over the last couple of years, it’s not likely that you’re doing things wrong in terms of SEO. Search is heading in a different direction and AI Overviews and AI Mode are directly answering the questions to almost any informational need."
In those cases, it has to use a page's publish date as a determining factor in ranking, with a median age of 200 days (compared to the 500 days noted above).
ahrefs.com/blog/why-cha...
9. The one place where age mattered less in this study was when it came to newsy content. Ahrefs explains this as AI's inability to differentiate relevance for news topics based solely on the page title and URL, because it doesn't have historical data to compare it to...
If we treat ChatGPT citations like ranking top-10 on a SERP, that's our comp. ChatGPT doesn't have something similar to Google's #1 spot, but it's worth noting the same Ahrefs study found the average age of the top SERP position was 5 years.
For context, a 2025 study by Ahrefs found that, if a net-new page was ultimately going to rank in the top 10 on Google, 40.82% of them would do so within the first month of being published...
In this research, Ahrefs found average page age was ~500 days, which is down form ~960 days in July of last year...
This is a departure from conventional SEO, where it's been a best practice to simplify URLs. For example "/dog-eye-blue" or "/xyz123" would be cited less on ChatGPT than, "/can-dogs-have-blue-eyes"
8. The average age of a page cited by ChatGPT is getting younger...
7. This is also really big. Ahrefs found that natural language URL slugs had 89.78% citation rate, versus 81.11% with less semantic value, like a code or non-natural language...
Note: A fanout query is ChatGPT's interpretation of the user's prompt. For example, a user may ask, "can a dog have blue eyes like a husky" and ChatGPT may internally change it to "Can dogs, like huskies, have blue eyes?"
If the platform feels the title and URL of another page is more relevant to its interpretation of the prompt, ChatGPT is going to cite the other page. Ahrefs found titles of cited pages had a moderate-to-strong cosine similarity to the fanout query (0.65).
Because ChatGPT isn't referencing their stored snippet when putting together citations for its response to your prompts, the title and URL of your page is doing all of the work...
Only about 2.5% of cited sources already had a snippet/summary created and stored in ChatGPT's system.
6. As Ahrefs says it, "Titles need to be semantically relevant to fan-out queries." This is the big one...
They'll actually open and crawl a page for that, likely in search of more specific information and added context.
5. Not having a "snippet" in ChatGPT's database (mostly) doesn't affect your likelihood of being cited...
Those only have a 1.93% citation rate, showing that the LLM likes to use Reddit as a source of information, but prefers to provide users with citations from other sources.
4. ChatGPT stores "snippet" summaries of pages, but doesn't use them when citing a page...
3. Of those 33 URLs, about 11 of them average out to be from Reddit. Despite the high Reddit volume, ChatGPT is less likely to cite from the pool of Reddit URLs...
If you rank well on Google, you're going to rank well on the AI platforms.
2. To form a response, ChatGPT pulls and average of ~33 URLs from the database. Think of this as a conventional Google user reviewing 33 blue links on a SERP before they decide what they want to click on and read.
Ahrefs did the SEO world a favor with a deep dive into ChatGPT citations. It's a long article, so here are my takeaways as a digital product manager overseeing user acquisition and technical SEO:
1. 88% of citations come from pages with high conventional search rankings. No huge surprise here... 🧵
Well...they say entrepreneurs should fall in love with the problem, not the product. www.adweek.com/brand-market...
Why talk about this today? Well, a fun story from TubeFilter that in a poll of TikTok users, 97% of them had followed a recipe they'd found on the app. How many of those recipes sucked? Hard to say. But it's probably fair to say the bad food influencers burned their cake, and ate it too.
If you mapped the level of social media influence someone has on a topic against the Dunning-Kruger Effect, you'd see their influence peaks at the same place their knowledge on the topic HITS THE FLOOR.
People who don't know how to cook are teaching people how to cook, feeding a death spiral of gross shit. Let's not be precious. It's gross.
The Alton Browns, Rachael Rays, and Gordon Ramsays of the world still have their place, but they now share it with the lady who dumps ice cream and boxed cake mix onto her hibachi grill. The result?
We can look at it across any genre in a long list of controversial topics: politics, science, religion...but it's more fun if we look at it through the lens of a less-controversial topic: FOOD INFLUENCERS.
The rise of de-influencing has only bolstered gains for people with trashy advice because the act of responding to a bad influencer only ups their relevance in the algorithm.
You want a long diatribe about food influencers? No? Too bad. Here's something I find fascinating about them.
The democratization of thought leadership through social media has destabilized the level of influence experts receive versus non-experts. 🧵⬇️
If you work in marketing, startups, or as a small business owner, let this be the single most important article you read this week. www.marketingbrew.com/stories/brea...
This is because bots are often doing real-time crawls of content when generating responses for users. They're trying to get to an answer as soon as possible, so they'll definitely err on the side of lazy and miss nuances in in-depth content. searchengineland.com/chatgpt-cita...
We're on the same page. *Most* organizations really shouldn't be looking for any future uptick in clicks and impressions. Those days are long gone. Consider goals that mitigate drops in those or shift to other metrics. searchengineland.com/retire-these...
We just launched a new product demo for our Business Intelligence dashboard. Forward us to a friend if you think they could benefit from some data-driven decision-making. In a few words: We deliver Big Data for Small Business. garvescope.com
