AI paid for by Ads – the GPT-4o mini inflection point

Original Article ↗ Hacker News Discussion ↗

Economics of AI-Generated Content + Ads

  • Many argue ultra-cheap models like GPT‑4o mini make ad-funded, auto-generated sites economically viable.
  • Others note the analysis ignores that content supply will explode, driving ad rates and revenues per page view toward zero.
  • Several point out arithmetic errors in the referenced revenue calculations, further undermining the article’s premise.
  • View that short-term arbitrage (making money “until everyone catches up”) has historically been common online, but not durable.

Impact on Search, SEO, and Content Quality

  • Widespread expectation of a “flood of crap content,” but many say the web was already saturated with SEO slop and content farms.
  • Concern that Google and others will increasingly answer queries directly with AI, cutting third-party sites out of ad revenue.
  • Some think Google is incentivized to punish low-quality AI sites; others say its results already surface low-value sludge.

User Behavior and Adaptation

  • Users already route around bad search (e.g., appending site:reddit.com or relying on niche communities).
  • Prediction that people will seek invite-only or smaller “human” spaces; non‑AI or “human‑only” content could become a selling point.
  • Skepticism that trust networks can scale; every trust system can be gamed or invaded.

AI Training on AI Output & Future Data Quality

  • Fears that future models will be trained mostly on LLM-generated text, causing gradual degradation (“telephone game”).
  • Counterpoint: serious model builders will curate and filter data, use human-labeled datasets, synthetic data with validation, and multi-model judging.
  • Some argue LLMs can, in principle, self-play or generate useful new combinations of ideas; others say they can’t truly create novel thoughts.

Value of Human Content

  • Split views:
    • One side: human-generated content will be drowned out, unfindable, and economically unviable.
    • Other side: many people write for passion, not money; human connection and authenticity will retain value, even if harder to discover.

Centralization, Control, and Ads in LLMs

  • Concern that AI answers will replace diverse links with a single, operator-controlled narrative, enabling powerful opinion shaping.
  • Expectation that free, centralized LLMs will eventually carry ads or sponsored answers; some foresee users moving to local/open models to avoid this.