Editorial Policy
This page states how AI Agent Hub creates, verifies, corrects, and monetizes its content. It exists so that the standards behind a published number are as inspectable as the number itself. The methodology page describes the review process in technical detail; this page states the policies that process operates under.
What we publish
The site publishes reference material on AI model pricing and agent engineering: verified rate tables, cost calculations with stated assumptions, technical guides, and evidence reviews of contested claims. Every indexed page carries a review date and a stable canonical URL. Pages that are incomplete, outdated, or awaiting source verification are excluded from the search sitemap and do not carry advertising code; the same applies to archived material.
We do not publish content whose only purpose is to attract search traffic. If a topic has nothing to add beyond what existing pages already establish — with sources — it is not published.
How sources are ranked
When a provider's own pages disagree, the page that determines billing outranks every other page. In practice this means the pricing or model documentation page outranks the launch announcement. This is not a theoretical rule: in September 2026 the launch post for a major model line and its own documentation pages disagreed by up to 5x on the same model, and the documentation rate is what appears on invoices. Where third-party figures are used — for example, a rate we cite from a tracker without independent re-verification — the page says so next to the figure. No number is presented as verified that we did not read from the primary source ourselves.
Data snapshots are dated. A rate checked on a given date is labeled with that date, and catalogue pages state when the whole set was last re-verified. Prices in this field change within weeks; an undated number is an unverifiable one.
Calculations and worked examples
Cost examples are calculations, not measurements, and are labeled as illustrative. Each states its assumptions — prefix size, step count, cache hit rate, peak or off-peak window — so a reader can substitute their own workload. Where a vendor publishes an estimate, we quote it as the vendor's estimate and show our own arithmetic separately. Where our independent calculation and a vendor claim diverge, both are shown. The interactive cost calculator exposes the same rates and assumptions it computes with.
Corrections
Errors are corrected in public. The correction record lists material corrections with dates — including a September 2026 catalogue re-check that fixed nine of ten stale price entries and a fix to reading-time labels that had overstated article length. To report an error, use the contact page; a correction request that cites a source gets a response and, if confirmed, a public record. We do not silently edit disputed claims.
Advertising and independence
The site displays advertising through Google AdSense. Advertising placement is handled programmatically and is not tied to editorial decisions in either direction: no provider pays for coverage, placement, or a favorable comparison; no page exists to promote an advertiser; we run no sponsored posts, paid reviews, or affiliate links in editorial content. Advertising code is excluded from archived and unindexed pages, and ad slots that cannot be filled degrade to internal links rather than empty boxes.
Advertising never appears inside data tables or calculation blocks. A reader can distinguish editorial content from advertising at a glance by placement and labeling.
How AI tools are used
AI Agent Hub is transparent that AI tools assist in drafting and maintaining this site. The workflow is: data is read from primary sources; drafts and calculations are prepared with AI assistance; every published figure is then checked against the primary source, and worked arithmetic is re-derived, before a page enters the indexed library. AI assistance accelerates production; it does not replace verification, and no number reaches a published page without passing the source check described above. This policy is itself an example of the standard: the claim is checkable against every dated data snapshot on the site.
What we do not do
- No paid placements, sponsored articles, or "advertorial" content of any kind.
- No publishing unverified leaks or single-source claims as fact; single-source figures are labeled as such or omitted.
- No benchmark scores presented as independent lab results when the source is vendor-published.
- No fabricated usage numbers, invented monthly bills, or traffic claims the site cannot measure.
- No scraping or republishing competitor content as our own.
Changes to this policy
Material changes to this policy will update the effective date above. Questions about any clause can be sent through the contact page.