39.6% of Developers Who Explain Why AI Replaced Their Searching Point to Cluttered, Ad-Heavy Search Results

Summary

Picture a bookkeeper who wants to know whether a client can deduct a home office. The search results open with ads, then a long article that buries the answer under a story. Next time, they type the question into an AI assistant. Developers on Hacker News have been describing that exact moment since late 2022.

We read every comment our phrase search found where someone said AI had taken over part of their searching, then looked at the 91 that said why. The top reason was not that AI is clever. It was that search results had become hard to use: 39.6% blamed ads, spam, SEO filler or falling quality.

Speed, a single direct answer and the ability to ask in plain language each came up in roughly a quarter of the comments. Over time, the reasons moved from escaping bad results toward liking what the AI gives back.

What we measured

We started from 198 Hacker News comments, posted from December 2022 to September 2026, that our team confirmed as real reports of someone using an AI assistant in place of a search engine. They came from 421 candidates found with phrase rules across 27.3 million live comments since January 2019.

Our team read each confirmed comment in full and coded every stated reason into one or more categories. A comment could have several reasons, so percentages add to more than 100. The 107 comments that gave no reason are left out of the shares. With 91 coded comments, each share carries a margin of error of up to about 10 percentage points at 95% confidence.

Cluttered results are the most common reason

People rarely described AI as perfect. Several said it still makes things up. They still preferred it to wading through pages built to rank and sell ads.

Typical complaints were recipe pages that hide the recipe, articles split across pages, cookie popups and result pages full of ads. The second group of reasons was about effort. AI gave one answer quickly, which saved opening several tabs and comparing them.

The reasons shifted from escaping clutter to liking the answers

We split the 91 comments into the first two years after ChatGPT's launch and the period since January 2025.

Early switchers were mostly pushed away from results they disliked. Later ones more often said they were pulled toward answers they trusted more, including answers with sources and up-to-date information. Both periods have small samples, so treat the size of the shift as a direction rather than a precise figure.

For coding questions, speed matters most

About a quarter of reason-giving comments (23 of 91) were about programming questions such as error messages, syntax or library examples.

Developers asking about code mainly wanted to save time and paste their own context. For everyday questions, the bigger draws were getting one clear answer and avoiding cluttered pages. Coding questions made up 38.8% of reason-giving comments in 2022 to 2024 (19 of 49) and 9.5% in 2025 and 2026 (4 of 42), as everyday questions took over.

What this means

The pages people say they are escaping are pages that make them hunt for the answer. If your service page, price list or how-to buries the answer under filler, an AI assistant may skip it and summarize a competitor who answered plainly. Put the direct answer near the top, keep facts like prices, hours and service areas current, and make each page easy to quote. Most of AI search optimization on an existing site comes down to edits like these.

AI assistants reward the same clarity people were asking for. Assistants can use some of your pages and not others, and a free AI visibility audit shows which; our generative engine optimization services then rewrite the ones where the answer is buried.

Limits

The sample is small. 91 comments gave a reason, so each share has a margin of error of up to about 10 percentage points, and the period and coding splits are wider still.

Reasons were coded by our team from what commenters wrote. Different coders could place some comments differently.

Comments were found with phrase rules, so switch stories worded in other ways were missed.

Hacker News commenters are mostly developers and founders. Their reasons may differ from those of the general public.

Dead and deleted comments carry no text in the dataset and are excluded. 2026 covers January to September.

Frequently asked questions

Related studies

• Developer conversation on Hacker News1 in 5 Popular Hacker News Stories Is Now About AI
• Developer conversation on Hacker NewsAI SEO Draws 3x the Positive Developer Comments of SEO
• Developer conversation on Hacker Newsllms.txt Splits Developers: 54% For, 46% Against
• Developer conversation on Hacker News'Written by AI' Comments on Hacker News Up 4.6x Since 2023