SEO in 2025: what still works, what doesn't, and what's next

Executive overview

Google's algorithm has undergone its biggest shift in a decade, moving from rewarding link-building to favouring established brands and high-authority domains. Small and niche sites that thrived on old-school SEO tactics are now volatile or penalised. The playbook has to be rebuilt from scratch.

Diversify across at least three channels — SEO alone is no longer a safe bet.

The algorithm has fundamentally changed

• Google shifted from link-based ranking to brand-preference ranking around 2022
• Major domains (top 100–500 sites) get near-automatic rankings regardless of content quality
• Indeed.com went from ~2M to ~20M monthly visitors publishing low-quality content at scale
• The old loophole was links; the new loophole is domain authority and brand signals
• If a site "smells like an SEO touched it," Google is increasingly likely to penalise it
• Old tactics — optimised anchor text, FAQ sections, internal linking formulas — likely hurt rankings now

Why affiliate SEO collapsed

• Niche affiliate sites boomed 2019–2022, then got caught in Google's crackdown
• AI-generated content flooded the web, forcing Google to respond aggressively
• Lars's affiliate network peaked at $7.2M annual revenue before collapsing
• Power-law dynamics apply: a handful of posts drive most revenue; one post made $100K/month
• High-value affiliate categories (VPNs, web hosting, credit cards, insurance) are widely known — no edge from discovery alone

What works now: the three-channel minimum

• SEO alone is too volatile; a 30% traffic drop is no longer worth investigating
• Minimum viable channel mix: one social platform (top of funnel) + SEO + email list from day one
• Google now weighs non-search traffic, branded keyword volume, and social presence as ranking signals
• Diversified brands rank more easily; SEO-dominant sites face constant volatility
• Niche SaaS with high ACV can still win on long-tail SEO — but keep SEO proportionate to other channels
• Avoid chasing keyword sets that fragment rapidly (local variants, celebrity net worth-style topics)

AI-generated content: why Lars says no

• AI produces "middle of the bell curve" content — adequate, but indistinguishable from competitors
• Cost to produce quality AI content (with skilled oversight) is roughly equal to hiring good writers
• The process of writing forces original thinking; AI removes that struggle and caps the output ceiling
• Rule for his freelancers: zero copy-paste from AI in final drafts; brainstorming and outlines are fine
• AI is useful for unblocking headlines or getting research starting points — not as a ghostwriter

AI overviews and the future of search traffic

• Google's AI Overviews (AIO) are answering queries directly, reducing click-through on informational content
• Content answering simple facts or definitions is most at risk of zero-click displacement
• Lars's hypothesis: absolute traffic will fall sharply, but conversion rates may rise — only high-intent readers will click through
• One reader found his HR site through ChatGPT — AI tools are already acting as referral sources
• The real risk: if Google deprioritises even high-quality niche content in favour of AI answers, the entire blog funnel shrinks

Parasite SEO explained

• Third-party SEO operators broker deals with major news publishers to publish content under their domain
• The operator runs thousands of SEO-optimised posts on a subsection of the publisher's site, piggybacking on domain authority
• Content quality is generally low; Google's domain-level preference is doing the ranking work
• Revenue from fully scaled parasite SEO operations: tens of millions per month
• Google has begun issuing manual penalties and policy changes targeting this practice
• Pattern is as old as SEO: Google signals a preference → SEOs exploit it → Google penalises → cycle repeats