The sources
What we base our work on
RankGrove borrows from research, and cites it with links. Here are the studies and documents the method is built on: what they actually say, and what we do with them in the product. Every source has been read and verified before making it onto this page.
Pew Research Center — searches with an AI answer (2025)
What it says: across 68,879 real searches by 900 adults tracked in March 2025, when an AI-generated answer appears, clicks through to websites drop from 15% to 8% of visits; the links cited inside the answer are clicked in 1% of cases; 26% of sessions end right there.
What we do with it: it is the whole reason GEO exists in RankGrove: visibility has to be built inside the answer, with content the AI can pick up.
“GEO: Generative Engine Optimization” — Princeton, KDD 2024
What it says: the academic study that defined the field: measurable optimisations, citations, statistics, structure, increase content visibility in generative engine answers by up to 40%, with different effects from sector to sector.
What we do with it: the shape of our articles, TL;DR, table of contents, tables, marked-up FAQs, sources, applies this evidence, article by article, automatically.
Google — “Rewarding high-quality content, however it is produced” (2023)
What it says: Google's official position on content generated with AI: ranking systems reward quality, originality and E-E-A-T, however it is produced. What counts is usefulness for people, whatever tool was used to build it.
What we do with it: it is the answer, with the primary source, to our customers' most common worry. We look after quality with validation checks before every delivery.
AI Act — Regulation (EU) 2024/1689, article 50
What it says: the European transparency obligations for content generated or manipulated by AI, applying from August 2026: marking in machine-readable format, and duties to inform the public.
What we do with it: compliance is built in: automatic marking, tracked approvals, a transparency note you can switch on. The details are on the page dedicated to the AI Act.
SparkToro — the zero-click study (2024)
What it says: for every 1,000 Google searches in the United States, only 374 clicks reach the open web; in Europe, 360. Most searches end inside Google, even before AI answers come into play.
What we do with it: the traffic that matters is earned by covering topics in depth, the logic of the topic tree, and by cultivating the channels where the answer is put together.
Our numbers, in public
To other people's research we add our own: the RankGrove blog is run entirely by RankGrove, plan, articles, images, publishing, and every month we publish the real numbers from the experiment in the “Data and experiments” category: impressions, clicks, citations. The first report arrives at the end of September 2026. Sources are borrowed; data, where possible, is produced.
This page is updated when new studies come out or when our own data adds something. The summaries are ours; the linked sources are what counts.
From theory to practice
All of this, applied to your blog
Fifteen days of trial: the method you have just read about, at work on your site.