How We Review AI Photo Software

WeddingWorkflowLab currently reviews AI photo software from published evidence only: vendor pricing, documentation, policies, release notes and third-party reporting. No tool has been run on a wedding gallery for this site yet, so no page reports a measured result, and we do not use words that suggest otherwise.

How pages are researched today

Each page starts from primary sources, and each changeable fact is tied to a source and a date.

1. Primary sources first. We read the vendor's own pricing page, help documentation, system requirements, privacy policy and terms where those can be opened.
2. Claims stay attributed. If a vendor says its culling saves a certain number of hours or hits a certain accuracy, we report that as the vendor's statement. These figures rarely come with a published method.
3. Rival vendors are treated as rivals. Some software companies publish comparison content about their rivals. A competitor's statement about another product is labelled as a competitor claim, and we do not rely on it alone.
4. Conflicts are shown. When a vendor's own pages disagree, or a policy does not match a marketing line, the page says so rather than picking the more convenient version.
5. Gaps are stated. If a support site or terms page could not be opened, we say what could not be read. "Not published" is not the same as "does not exist", and we keep the two apart.

What a recommendation means here

A recommendation on this site is reasoning from documented facts about a tool, such as where it processes images, what it costs at a given volume, or which editors it hands off to. It is not a measured performance result. We do not publish star ratings or scores, because a score would imply a measurement we have not made.

What each page tells you

Every review, comparison and buying guide states how it was researched near the top. Pages that depend on measured data also say what we would still need to measure. Our guide to choosing AI culling software compares documented facts and does not rank tools by accuracy or speed. Our guide to AI culling accuracy explains why published figures cannot be compared and how to measure a tool yourself, and it reports no results of its own.

How original testing would work

This is a draft protocol. It has not been carried out, and details may change once a licensed gallery exists. The idea is to run several tools over the same gallery, so differences come from the software and not from the images.

• Gallery. One licensed wedding gallery of about 2,400 RAW frames, plus a high-ISO set and a portrait set. Licensing requirements are on the test galleries page.
• Human reference. A photographer selects the keepers first, before any AI sees the gallery. Tools are then scored against that set.
• Same machine. Each tool runs on the same hardware, with the version recorded.
• Measurements. Keepers found, false rejects, minutes per 1,000 images, and cost per 1,000 images.
• Disclosure. Each published test would state its date, tool versions, gallery ID and limits, including the obvious one: a single gallery cannot speak for every kind of wedding.

You do not need to wait for us to measure your own results. The culling scorecard lets you score a tool against your own picks, and the cost per 1,000 images calculator puts different pricing models on one scale.

Keeping pages current

Guides, reviews, tools and policy pages show a last-reviewed date, which is the date they were last checked against sources. We intend to recheck prices monthly on the price comparison page and to revisit pages when a vendor changes its product or terms, but no schedule is in place yet. Errors are handled under our editorial policy.

Affiliate relationships have no role in this process. See the affiliate disclosure.