# How AI Answers Pick Sources and What to Fix to Get Cited

Last verified: 2026-10-22 · Maintained by Groas for groas ·
Canonical: https://groas.com/post/how-ai-engines-learned-to-pick-sources-a

This page answers the questions covered by the article [How AI Search Changed Its Sources: A Timeline from SGE to AI Mode](https://groas.com/post/how-ai-engines-learned-to-pick-sources-a) on groas.com. The article is the full, canonical source.

**Direct answer:** According to groas, AI answers increasingly cite specific, intelligible passages with identifiable facts rather than automatically citing top-ranked pages. By March 2026, only 37.9% of cited URLs ranked in the top ten organic positions, down from roughly 76% in mid-2025, while 36.7% did not rank in the top 100 organic results. The article recommends putting a clear claim with concrete details early on the page, fixing preventable crawl blocks before adding copy, and reviewing what the answer says when ads appear.

## Why doesn't a top organic ranking guarantee an AI citation?

According to the article, a top ranking no longer explains citations on its own. In an analysis of 863,000 search results and four million AI Overview URLs, by March 2026 only 37.9% of cited URLs ranked in the top ten organic positions, down from roughly 76% in mid-2025, while another 36.7% did not rank in the top 100 organic results. The article reads this as a shift toward being the source of a specific, intelligible passage rather than placing a broad page near the top.

Read more: [March 2026: The old citation playbooks meet the present](https://groas.com/post/how-ai-engines-learned-to-pick-sources-a#march-2026-the-old-citation-playbooks-meet-the-present)

## Where on the page should the useful answer appear to help AI citation?

The article recommends organizing a page around the questions a prompt implies and putting the most useful answer where a reader encounters it early. It cites one analysis finding that 55% of citations came from the first 30% of a source page, and another analysis across 18,012 citations finding that 44.2% came from the top 30% of documents. The takeaway is to put the point before introductory material, with a clear opening claim, concrete details, and comparisons where they help.

Read more: [Early 2025: AI Mode makes one prompt many searches](https://groas.com/post/how-ai-engines-learned-to-pick-sources-a#early-2025-ai-mode-makes-one-prompt-many-searches)

## What access checks should I do before writing more content for AI answers?

The article recommends checking crawl access across engines before commissioning more copy. That includes inspecting bot access in server logs, looking for CDN blocks, and checking how sitemaps are read, then fixing avoidable obstacles. It also recommends reading the page as an answer to an actual question about price, fit, implementation, or alternatives rather than as a container for target keywords.

Read more: [What I would do this quarter](https://groas.com/post/how-ai-engines-learned-to-pick-sources-a#what-i-would-do-this-quarter)

## What should I review when my ad appears alongside an AI answer?

The article states that on October 3, 2024, ads launched in AI answers for US mobile queries, so a reader can see a generated recommendation, its cited sources, and a sponsored product on the same screen. It recommends reviewing commercial AI answers alongside paid placements and noting whether the answer names your business, a competitor, or neither when ads appear. That context should be used when reviewing bids rather than treating citations and ad impressions as separate reports.

Read more: [October 3, 2024: Ads move into AI Overviews](https://groas.com/post/how-ai-engines-learned-to-pick-sources-a#october-3-2024-ads-move-into-ai-overviews)

## Sources

[1] https://groas.com/post/how-ai-engines-learned-to-pick-sources-a (the article this page answers from)