SEO advice is full of numbers nobody can trace. Here we collect the research that holds up: academic papers, independent studies, controlled tests and official case studies. For each one we checked the figures against the source, then wrote down what it found, what to do about it, and what to keep in mind before you quote it.
18 pieces of research · 7 themes · every figure checked on October 13, 2026
The big picture
AI search and clicks
AI answers take clicks on long, informational queries, and being cited inside the answer is now what separates a small loss from a large one. The routes in are ordinary SEO plus content that is easy to quote: specific numbers, named sources, clear statements.
- AI features and your website (Google's own guidance): Google Search Central, 2025. Google says there are no extra requirements and no special optimizations needed to appear in AI Overviews or AI Mode. A page must be indexed and eligible to show a snippet.
- AI Overviews and click-through rate: 15 months of client data: Seer Interactive, 2025. Organic CTR on queries with an AI Overview fell 61% over the period, from 1.76% to 0.61%.
- Google users click less when an AI summary appears: Pew Research Center, 2025. People clicked a regular result on 8% of visits to pages with an AI summary, against 15% when there was no summary.
- AI Overviews and the top-ranking page: 300,000 keywords: Ahrefs, 2025. The top-ranking page got about 34.5% lower average CTR when an AI Overview was present than the control group trend would predict.
- GEO: Generative Engine Optimization: arXiv (IIT Delhi, Princeton University and independent researchers), 2023. Adding quotations from relevant sources raised visibility the most, about 41% over the unchanged page on their main measure.
Clicks and zero-click search
Most searches already ended without a click before AI Overviews. Measure impressions and brand demand alongside clicks, and treat YouTube and Maps as part of search.
- Zero-click search in 2024: most Google searches end without a click: SparkToro (data from Datos), 2024. 58.5% of U.S. Google searches and 59.7% of EU searches ended without a click on any result.
Links and authority
Links still separate pages that get traffic from the 96% that get none, and the idea behind them has not changed since 1998: links from well-linked pages, with descriptive text, count most.
- 96.55% of pages get no traffic from Google: Ahrefs, 2023. 96.55% of pages got no traffic from Google, and 1.94% got between one and ten visits a month.
- 11.8 million Google results: what correlated with ranking: Backlinko, 2016. The top result had about 3.8 times more backlinks than results in positions 2 to 10, and about 3 times more referring domains.
- The Anatomy of a Large-Scale Hypertextual Web Search Engine: Stanford University, WWW7 conference, 1998. PageRank scores a page by the scores of the pages linking to it, with each linking page's score shared across its outgoing links.
Content quality
Google's own checklist and independent research point the same way: original information, clear authorship and methods, and no mass-produced pages. Spam tactics that win briefly tend to lose at the next update.
- Creating helpful, reliable, people-first content (Google's guidance): Google Search Central, 2026. Google asks whether content adds original information, research or analysis, and goes beyond the obvious.
- Is Google getting worse? A year of product review search results: ECIR 2024 (Leipzig University, Bauhaus-Universität Weimar), 2024. Affiliate pages are rare on the web overall (about 2.35%) but common in review results: 29% on Google (via Startpage), 42% on Bing, 41% on DuckDuckGo.
Speed and page experience
Speed work pays off most clearly in conversions and engagement. Measure it in revenue with an A/B test where you can.
- Core Web Vitals business results from 19 companies: web.dev (Google), 2021. Tokopedia improved LCP 55% and saw 23% longer average sessions.
- Vodafone: 31% faster LCP, 8% more sales: web.dev (Google), 2021. Largest Contentful Paint was 31% better on the optimized page (5.7 seconds against 8.3).
Structured data and rich results
Structured data helps most where it earns a visual result (recipes, video, jobs) and after the technical basics are fixed.
- MX Player: 3x Google traffic with video markup and sitemaps: Google Search Central, 2021. Traffic from Google grew 3 times.
- Saramin: doubling job-season search traffic with technical clean-up: Google Search Central, 2020. Organic traffic rose 15% in the first year from fixing crawl errors alone.
- Rakuten Recipe: 2.7x search traffic with recipe structured data: Google Search Central, 2018. Search engine traffic to recipe pages rose 2.7 times.
SEO testing
Template changes on big sites can move traffic 10% or more either way, and "best practice" sometimes loses. Test before you roll out.
- Removing SEO text from category pages increased mobile traffic: SearchPilot, 2024. Removing the text produced a statistically significant increase in organic traffic from mobile devices.
- 10 SEO A/B tests that moved organic traffic by 10% or more: SearchPilot, 2024. Moving the brand name to the front of title tags lifted a retailer's organic traffic over 15%.
Where the research disagrees
- Are AI Overview clicks good or bad? Google says clicks from pages with AI Overviews tend to be higher quality. Pew, Seer and Ahrefs all measured fewer clicks. Both can be true: fewer visits, but the ones that come may be more engaged. Judge it on your own conversions, not on either claim.
- Is long "SEO text" helpful? Google says it has no preferred word count, and a SearchPilot test found that removing category page SEO text increased mobile traffic. A 2016 correlation study found no link between word count and ranking. Topic coverage matters; length for its own sake does not.
How we pick and check research
- We prefer studies that state their sample, method and dates, and controlled tests over correlations.
- We read every source ourselves and check each figure against it. If we cannot read the source, it is not here.
- We say who published it and whether they sell something related, because many SEO studies come from tool vendors.
- Our summaries are our own words. Follow the source link for the full study.
See also our own original research: old domains, expired domains and keyword difficulty.