In August 2026, Google changed how AI Overviews display citations. Instead of a source list at the bottom of the card, citations now appear inline, as numbered links embedded directly in the AI-generated text, next to the claim they support. Early data from publishers who shared Search Console numbers shows click-through rates roughly quadrupled compared to the old bottom-only format.
If you were optimizing for the old citation format, the target moved.
What changed
Before August 2026, AI Overviews displayed a text summary with a carousel or list of source links at the bottom. Users had to scroll past the AI text to find the sources. The Nielsen Norman Group had previously found that users exhibit "satisficing" behavior in search: they stop scanning once they find an acceptable answer. When all citations lived at the bottom, source pages received minimal traffic.
The new format places clickable references directly in the text. When the AI Overview makes a specific claim, a numbered link appears adjacent to that statement. Bottom citations still exist but serve as supplementary "additional sources" rather than the primary attribution mechanism.
First Page Sage's CTR data confirmed the shift: citation CTR roughly quadrupled between December 2025 and August 2026, coinciding with the inline format rollout.
The numbers
A study of 6,200 AI Overviews captured between August 12 and 28, 2026 measured citation placement patterns and early CTR data from 23 publishers who shared anonymized Search Console data:
| Metric | Inline citations | Bottom-only citations |
|---|---|---|
| Median CTR | 3.1% | 1.1% |
| CTR range | 2.4% to 3.8% | 0.8% to 1.4% |
| Lift | 2.8x | baseline |
For comparison, traditional blue links in positions 1-3 below the AI Overview still get 4.2% to 8.7% median CTR. Inline citations in AI Overviews are now closer to that range than to the old bottom-only format.
What content gets inline placement
The study also broke down CTR by content type:
| Content type cited | Inline CTR | Bottom-only CTR | CTR lift |
|---|---|---|---|
| Statistical data and research | 4.3% | 1.4% | 207% |
| How-to instructions | 3.8% | 1.3% | 192% |
| Product comparisons | 3.2% | 1.0% | 220% |
| Definitions and explanations | 2.1% | 0.9% | 133% |
| News and current events | 2.9% | 1.1% | 164% |
Statistical data and research content gets the highest inline CTR (4.3%) and the largest lift over bottom-only placement (207%). Pages with specific statistics, original research data, and factual claims that need sourcing are the most likely to earn inline placement and the most likely to get clicked when they do.
67% of inline citations in research-oriented queries pointed to original studies rather than secondary coverage. AI Overviews preferentially cite the source of data, not articles summarizing that data.
What this means for your content
Three concrete changes:
1. Put statistics and factual claims near the top of your content. The Babel42 study (October 2026) found that 67% of AI Overview quoted passages come from the first 30% of the page. If your key statistics are buried in the conclusion, they are less likely to be extracted. If they are in the first two sections, they are more likely to become an inline citation.
2. Cite your own sources inline. Pages that link to primary sources (studies, official documentation, government data) are more likely to be positioned as the "source of data" rather than a secondary summary. AI Overviews cite the source, not the article about the source. If you are the page that links to the original study, you become the intermediary. If you are the original study, you are the citation.
3. Structure factual claims as self-contained sentences. Inline citations attach to specific claims. A sentence like "78% of AI Overviews cite pages outside the top 10 organic results" is easy to attach a source link to. A paragraph of context without a discrete factual claim is harder to cite inline.
The limitation
The CTR data comes from 23 publishers who shared anonymized Search Console data. That is a small sample, and publishers who share data may skew toward those with better-than-average performance. The 2.8x lift is a median, not a guarantee. Individual results will vary by industry, query type, and position within the AI Overview.
The format change is also still rolling out. Some queries may still show the old bottom-only format. Google does not publish which queries use which format.
What to check
If you want to see how your pages score on the signals that correlate with AI Overview citation, run a free audit on geoready.dev. The audit checks content structure, schema markup, published dates, named authors, and canonical URLs, all signals the WashU study found associated with AI Overview citation. The same engine is available as a CLI on GitHub, open source, MIT licensed.
The audit tells you if your pages are structured for citation. It does not tell you if Google will cite them inline. But if your pages are missing the structural signals, inline placement is not even in play.
Top comments (0)