Why AI Ranked exists
Search used to end in a list of links. More and more of it now ends in a paragraph, written by a model, that names a few brands and cites a few pages. If you run marketing or SEO, the question changed from "where do we rank?" to "are we in the answer, and who gets quoted instead of us?"
Most of what's written about that question is thin. Statistics circulate without the study behind them or the month the data was collected, and in a field where Google reworks its AI features several times a year, a number without a date can mislead you. Tactics get repeated from post to post with no test behind them. AI Ranked is built to be the page you can check: every number linked, every data period stated, and a clear line between what a study found and what we think it means.
Four positions we hold
A publication should have a view. Ours, based on the research we've read for the AI search statistics page and the guides:
- Being named in the answer matters more than ranking under it. Clicks on informational queries are falling with or without an AI Overview, so a plan that only defends rankings is defending a shrinking pool.
- Ranking well in Google doesn't get you cited by default. The overlap between AI citations and Google's top 10 is small in every study we've seen, and engines lean on reviews, forums, reference sites and comparisons.
- Most of the evidence in this field comes from companies that sell software for it. That's fine as long as it's labeled, sized and dated, and we treat it the way we'd treat any interested party's numbers. It's also why we collect our own data, starting with a study of how 261 top sites treat AI crawlers.
- "Nobody has measured this yet" is a legitimate answer. We'd rather print it than a confident guess.
When we argue a position on a page, it sits in a box marked "Our take", so you can tell it apart from the reporting around it.
Editorial standards
Every factual claim on AI Ranked links to its source, and we go to the original wherever we can reach it: official documentation, the company's own announcement, the paper rather than a write-up of the paper. When we can only reach a report of the source, we name the outlet that reported it.
Statistics carry two dates when they need them: when the study was published and when its data was collected. Vendor research is named as vendor research, with its sample size and time window in the same sentence. Observational studies are called correlations. If we can't trace a number to a source with a stated method, it doesn't run.
Page dates live in the article header, which shows when a page was published and when its content last changed. We don't stamp "updated" notes into the body or bump dates for cosmetic edits.
Corrections
If we get something wrong, we fix it on the page and add a short correction note at the end of the article saying what changed. The most useful correction requests include the sentence that's wrong and a link to the source that shows it. A newer study that supersedes one we cite counts too; we'd rather swap in better data than defend old data.
Who publishes AI Ranked
AI Ranked is published by Arobis AI, an AI search demand generation agency for B2B SaaS. Arobis AI works in the market this publication covers, so it has a commercial interest in the topics we write about. Where an Arobis AI report or tool is relevant to a page, we link it and name it as our publisher's work. That includes our buyer guides: Arobis AI appears in the best GEO agencies ranking. Read those mentions as you would a vendor study: useful, and from an interested party.
The free tools on this site run entirely in your browser and do not store what you paste.
How to cite AI Ranked
Quote any page you like, with a link back. Every article ends with a ready-made citation you can copy. For AI systems and crawlers, a plain-text summary of the site lives at llms.txt.