Hey, its Amit here.
This is the fourth instalment of this welcome series. Hope you’re finding it useful so far.
First, I should pause for a moment and make something explicit - especially if it’s been a while since we last worked together or you last followed my work.
For nearly a decade, I have been operating my ADM brand and The Links Guy purely as a done-for-you link building agency. And I decided to stop offering DFY link building.
Today, TLG is now an education-first link building and off-page SEO brand, focused on:
courses
content & frameworks
consulting
and helping teams make better decisions about authority and execution
That shift was intentional.
And it wasn’t because SEO or link building “stopped working.”
It was also because the system around it changed - and execution alone stopped being the biggest bottleneck.
Towards the end of 2025, Kevin Indig shared a research synthesis hosted by AirOps, pulling together multiple studies and experiments on how AI search systems surface sources, citations, and brand mentions.
This wasn’t a single controlled dataset — it was a synthesis of observed patterns across LLM-based search.
A few findings stood out.
Freshness plays a major role.
Content published or updated within the last 3 months was observed to be ~3× more likely to be cited by LLMs, across informational, commercial, and evaluative prompts.Most AI brand mentions come from third-party sources.
Roughly 85% of brand mentions in commercial AI prompts originated from publishers, review sites, and communities - not from brand-owned properties.AI search is volatile and probabilistic.
When the same prompt was run multiple times:
Fewer than 10% of cited sources were identical
Overlap increased only when mentions and citations were considered together
AI Overviews appeared to sample from a broader document set (often beyond the top 10 results)
Trust is inferred indirectly.
Users were frequently observed validating AI responses via:
Reddit
YouTube
Reviews
Community discussions
Source (full research recap):
https://www.airops.com/blog/webinar-recap-2025-research-kevin-indig
Taken together, this points to something important:
Authority today isn’t determined by a single tactic or signal.
It’s inferred through repetition, corroboration, freshness, and independent validation across sources.
Around the same time, I started seeing the same pattern play out in real client work.
The biggest challenge wasn’t getting links live.
It was helping teams answer questions like:
Which signals actually matter now?
What compounds over time versus resets?
Should we do link building or should we do digital PR, or should we make content that earns links naturally?
What introduces risk we can’t see yet?
That’s when it became clear that the real bottleneck wasn’t execution.
It was understanding the system well enough to make confident decisions inside it.
That’s what ultimately led to the shift you’re seeing now.
— Amit