AI is killing All About Berlin
TL;DR Highlight
After Google AI Overviews, a Berlin guide site lost 70-75% of its traffic.
Who Should Read
A developer running a content or docs site that depends on search traffic and worries about AI summaries.
Core Mechanics
- Traffic fell 70-75%
- All About Berlin, run since 2017, grew for 7 years and then reportedly lost 70-75% of its traffic to Google AI Overviews.
- Search used to put a link to the site first.
- Result order changed
- Now users see ads, then an AI answer trained on the author's content, then finally the link.
- The author calls this devastating for traffic.
- It's not only about money
- The author put 8 years into the guides and feels he loses his voice and audience when they are blended into AI answers.
- Government offices refer people to the guides, the Berlin immigration office uses his data, and it inspired two government portals.
- Other sites are in the same spot
- Many site owners reached out privately.
- Some paywalled their work, others keep sites online but stopped updating them.
- He expects it to get worse
- The author says Google announced last week that AI answers will replace search results entirely and will include ads.
- He describes this as a future with no websites, just a chat box with ads.
Community Reactions
- AI summaries are like ad blocking, one commenter argues
- The comment says HN's long-held stance that the user agent serves the user's wishes is indistinguishable from AI summaries.
- It argues users clearly want AI Overviews.
- AI kills its own source, some commenters say
- If traffic dies, sites die and the AI has nothing to pull from.
- Fewer clicks also cut Google's own ad revenue.
- Counterpoint on the affiliate business model
- One commenter criticized the site for recommending affiliate-link banks first.
- They argue a business that depends on a third party it can't control was never a good one.
- GEO won't save it, one commenter says
- Even when cited, almost nobody clicks the link.
- Being the AI's source and being a business are now separate things.
How to Apply
- Measure your search dependence first
- Today, check what share of your total traffic comes from Google search in your analytics.
- Pull monthly trends for the last 1-2 years and look for sharp drops.
- Build a direct channel to readers
- Open one channel such as a newsletter or an app that connects you directly to readers.
- Commenters mentioned moving to a newsletter or an app.
- Try the AI bot blocking setting
- Check Cloudflare's AI bot blocking setting and turn it on if needed.
- A commenter says Google respects the opt-out for AI training.
Terminology
Related Papers
Gemini 4 Neon
Gemini 4 Neon cuts cost per task 32% versus Carbon, at $3 input and $15 output per million tokens
Not Everything Needs an LLM
PostHog used small decision models to skip 98.7% of expensive LLM calls
GPT‑6.2 Astra and Sol
GPT-6.2 Astra and Sol cut costs sharply, but over-engineering and early-stop reports are mixed
Show HN: TrnRT – An open-source LLM inference engine for AWS Trainium
One trn2.48xlarge serves GLM-5.3-Flash through an OpenAI-compatible API using NKI megakernels.
Know When to Zoom: How Small Is Too Small for Vision-Language Models?
Splitting large photos into tiles raised Claude and Gemini accuracy by 11.1–14.5 points
Understanding Positional Embeddings in Modern LLMs
A walkthrough of how LLMs encode token position, from GPT-2 to RoPE
Show HN: DeepSeek V3 in ~200 lines of pure Jax
DeepSeek V3's core design in 190 lines of JAX, trained on a MacBook Air in 2.6 hours
Nvidia allegedly engineered scarcity by funding AI startups to buy its GPUs
WSJ alleges NVIDIA funded startups that then bought its own GPUs to prop up demand.