Methodology
How we know what we publish.
AIRecruiter.co is an independent research firm for the talent economy. Our only asset is trust, and trust is earned one number at a time. This page is the standard behind every figure we publish: where the data comes from, how we label our confidence in it, how we reconcile conflicting reports into a single defensible statistic, who does what between our automated ingestion and our human analysts, and how we handle corrections. If you can read only one page before citing us, read this one.
- Cited sources
- 190
- Statistics
- 107
- Trackers
- 6
- Tracker window
- 2000–2026
01The sourcing standard
Every number resolves to a primary source.
The rule is absolute: no statistic exists in the Index without at least one real, resolvable source URL. We do not publish a figure attributed to a source name alone. “According to industry research” is not a citation. Every number we ship links to the specific page, dataset, or filing where a reader can pull the identical figure themselves. Today the Index rests on 190 cited sources, spanning government labor data, primary industry research, filings, and vendor disclosures, and every one of them is listed in our public source registry.
The 6 trackers hold themselves to the strictest version of this standard. All of them are official U.S. Bureau of Labor Statistics series, retrieved from FRED (the Federal Reserve Bank of St. Louis), covering 2000–2026 and measured the same way every year. Nothing in a tracker is estimated, smoothed, or hand-set, and each one links to its FRED series page so any reader can download the exact numbers. Where a figure is the latest available reading rather than a completed annual average, we flag it as partial rather than presenting it as final. Showing our work is not a courtesy. It is the product.
02Confidence tiers
We separate what is measured from what is estimated.
Not all data is equally certain, and pretending otherwise is the fastest way to lose a reader’s trust. So every figure in the Index carries one of three honest confidence labels, and we never inflate one. Across 107 statistics, the balance is deliberately conservative: the majority is measured or credibly reported, and the small modeled set is always flagged as such.
Verified
42Confirmed against the primary source.
We opened the source and read the number ourselves. All 6 trackers are verified because they are official, measured government series.
Reported
52Consistently reported by a credible publisher.
The figure is consistently reported by a credible publisher, but is not something we independently measured against a primary instrument.
Modeled estimate
13A market estimate or projection, not a measured figure.
A market estimate or projection, not a measured count, for example a forecast of market size or a projected share for a future year.
A reader should be able to sort our entire library by how much weight a number can bear, because we tell them plainly. You can filter the whole Index by tier in the statistics explorer.
03Reconciliation
Conflicting figures become one canonical statistic.
The talent market is measured by dozens of publishers using different samples, definitions, and time windows, so the same question often has several plausible answers. Our job is not to list all of them. It is to resolve them into one defensible, apples-to-apples statistic and to be explicit about the nuance we set aside. When figures conflict, we identify what each one actually measures, choose the most rigorous and directly comparable primary source, publish that as the canonical number, and record the surrounding nuance in the figure’s context and trend notes and in the deep-dive report behind it. We do not publish near-duplicates that only look like disagreements.
Worked example
How many organizations use AI?
What a reader finds in the wild
- 43%of organizations use AI in HR (SHRM, 2025)
- 51%use AI specifically to support recruiting (SHRM)
- 69%of HR professionals report using AI for recruiting
- 37%of TA professionals use generative AI daily (SSR)
- 18%use AI broadly across hiring (SSR)
Each figure is true of something different: a different population, a different definition of “use,” and a different scope. The distinctions are not noise. They are the story.
What we publish
Organizations using AI in HR
Nearly half of organizations now use AI for HR tasks, a sharp acceleration in mainstream adoption.
Up from 26% in 2024, a 65% year-over-year increase.
We publish the single defensible number a reader can cite. The rest lives as context and in the deep-dive report as the broad-but-shallow story: high penetration, low depth.
04Division of labor
Machines ingest at scale. Analysts verify and frame.
We are an AI-native firm, and we are precise about what that means. AI lets us cover the entire talent market at a scale no traditional research shop could staff. But no number reaches a reader on the strength of ingestion alone.
Automated: ingestion
We continuously pull funding rounds, acquisitions, product launches, leadership moves, filings, and published research across the whole ecosystem, then extract candidate figures from that flood. This is where the machines earn their keep.
Human: verification, labeling, reconciliation, framing
A human analyst confirms each figure against its primary source, assigns its confidence label, decides which figure is canonical when sources conflict, writes the framing that tells a talent leader what it means, and wires it up to its source and down to the report that establishes it. Every time.
The subject of our work is always the talent market, never the technology we use to study it.
05Corrections and updates
We date our data and we publish corrections.
A figure without a date is a rumor, so every number we publish carries the period it refers to. Each statistic states its own as-of date, the trackers are labeled with their measurement window and latest reading, and the current-year tracker point is marked partial until the full-year average is final.
When a primary source revises a number, or when we find that we got one wrong, we correct it rather than quietly deleting it. If a correction changes the substance of a published finding, we note what changed and when. A firm that never appears to be wrong is a firm that is hiding its mistakes. Being correctable in public is a feature of trustworthy research, not an embarrassment to it.
06Independence
No vendor pays for coverage.
We map and analyze the companies that sell recruiting technology, and none of them pays us to do it. We take no placement fees, no listing fees, and no sponsorship from the vendors we cover. Where a vendor appears in our market map or a company figure appears in the Index, its position reflects our analysts’ read of the evidence and nothing else.
We are not a recruiting tool, we do not place candidates, and we have no stake in which vendor wins. That is the entire reason a head of talent, an operator, or an investor can use our research to make a decision: the source has no side in the outcome.
07Citation
How to cite AIRecruiter.co.
Our research is built to be quoted, and we would rather you cite us precisely than paraphrase us loosely. When you use a number from us, link to it and carry its confidence label and as-of date with it, because those are what make it defensible in your own work. Every figure in the Index links up to its primary source and down to the report that establishes it, so a reader can always trace a claim to the ground it stands on.
The Index
The AIRecruiter.co Index. AIRecruiter.co, accessed [date accessed]. https://airecruiter.co/data
A single statistic
AIRecruiter.co, "Organizations using AI in HR" (43%, as of 2025, Verified). The AIRecruiter.co Index. Primary source: SHRM, 2025, https://www.shrm.org/topics-tools/research/2025-talent-trends/ai-in-hr. https://airecruiter.co/data
A deep-dive report
AIRecruiter.co Research (2026). The State of AI in Recruiting 2026. AIRecruiter.co. https://airecruiter.co/blog/state-of-ai-recruiting-2026
Building on our data programmatically? The whole Index is available as a machine-readable JSON export, with every statistic’s value, confidence tier, as-of date, and source URL.
Now read the research the method produces.
Explore the Index, or read the deep-dive reports where every figure is established and contextualized in full.