Technology

First published

AWS lists a website check at one US cent. What does it cost to confirm a barrier?

Our interpretation

If you run a website, compare what it costs to confirm a real barrier, not just the cost of a scan. Siteimprove lists a check at US$0.01 and a suggested fix at US$0.13 through AWS. Neither price tells you whether a reader can finish a task. Before changing what you buy, use the same pages and count only barriers a person can reproduce. Keep human review in the budget.

The Signal in brief

What changed

AWS Marketplace now lists per-request billing: US$0.01 per detection call and US$0.13 per AI remediation call. The purchase route is new; accessibility scanning was already available.

Why it matters

A cheap request is not necessarily a cheap confirmed fix.

Who it affects

Website editors, accessibility leads and small digital teams.

Where

AWS Marketplace, US dollar billing for compatible agent setups with approved payments. Local purchase eligibility not verified.

What to watch

Check the current quote and your own bill before relying on estimates. Compare reproducible barriers from each method and retest any fix separately.

No live transaction, invoice, automation quality, successful remediation or compliance verified.

Explore this signal

Emerging data

A new way to pay, not a new scanner

Siteimprove says eligible teams can now pay per request through Amazon Bedrock AgentCore, an AWS service for running AI agents. Charges appear on the AWS bill and are subject to the organisation’s buying settings. In June the vendor already offered accessibility checks and suggested fixes through its MCP server, a way for software to call those tools. Its August help described a subscription route. The new evidence is a payment option, not better detection.

A new way to pay, not a new scanner
  1. 30 Jun 2026Accessibility MCP tooling announced
  2. 25 Aug 2026Help documentation updated; prior subscription route described
  3. 20 Nov 2026 SGTPer-request marketplace option announced
Tool functionality and billing changed at different times.
Sources
  • Siteimprove issuer wire · 2026-11-19
  • AWS Marketplace listing
  • June Siteimprove MCP launch · 2026-06-30
  • August Siteimprove help · 2026-08-25

Check what a reader can do

On the same set of public pages you control, use a keyboard to repeat a reader task such as submitting a form. Compare your existing check with the new route. Count distinct barriers a reviewer can reproduce. A flag is not a confirmed barrier, and a suggested fix is not a tested change.

Check what a reader can do
  1. 1Pick one reader task on matched pages
  2. 2Reproduce distinct barriers manually
  3. 3Compare full costs against your existing check at like-for-like coverage
Proposed comparison, not observed performance.

Count the whole review

At the listed rates, 100 detection calls cost US$1 and 10 suggested-fix calls cost US$1.30. At an assumed 30 minutes of review valued at US$40 an hour, staff time adds US$20. The US$22.30 total is illustrative, not a quote or invoice. It excludes tax, other platform costs and retesting.

Count the whole review
  • 100 checks × US$0.01US$1.00 at listed detection rate
  • 10 suggested fixes × US$0.13US$1.30 at listed remediation rate
  • 0.5 staff hour × US$40US$20.00 assumed staff time
  • Partial illustrative totalUS$22.30 before tax, other charges and retesting
Not a quote or invoice.
Sources
  • AWS Marketplace listing

Report this passage

Research checked at

The connection

June tools and August subscription instructions show capability predates November billing change.

Report this passage

The opportunity

Agree one reader task, same public test pages and a human check. Use existing approved access or request a quote; do not enable spending merely to test.

What could happen next

  1. First-order consequence

    The listing advertises a new buying route.

  2. Second-order consequence

    A team with sporadic needs may compare a per-call quote with existing coverage.

  3. Third-order consequence

    Extra flags may create review work; count only reproducible new barriers.

  4. Fourth-order consequence

    If a fix ships, retest the original task, including keyboard use.

  5. Fifth-order consequence

    If useful repeat results cost less at equal coverage, change the review mix. No such outcome established.

Costs, benefits and assumptions

At the listed rates, 100 detection calls cost US$1 and 10 suggested-fix calls cost US$1.30. At an assumed 30 minutes of review valued at US$40 an hour, staff time adds US$20. The US$22.30 total is illustrative, not a quote or invoice. It excludes tax, other platform costs and retesting.

What would change our view?

June MCP checks already existed; the new payment route can duplicate coverage teams already have.

Alternative explanation

Existing subscription or manual checks may be sufficient, especially when reviewer time drives the bill.

What would confirm it

Same pages and task, lower all-in cost per reproducible barrier at like-for-like coverage in repeated reviews; successful remediation requires a separate retest.

What would disprove it

No incremental reproducible barrier, worse task coverage, or review cost erases the price difference. Zero barriers means do not divide by zero or declare a saving.

Evidence and sources

  • Siteimprove issuer wire

    Publisher: Siteimprove

    Original language: English

    published date precision: day

    retrieved date precision: day

    Source publication date 2026-11-19

    Source retrieval date 2026-11-20

    Coverage and independence

    19 Nov 17:06 EST = 20 Nov 06:06 SGT. New per-request AWS billing, not a first accessibility scanner.

    View source
  • AWS Marketplace listing

    Publisher: AWS Marketplace product listing

    Original language: English

    retrieved date precision: day

    Source retrieval date 2026-11-20

    Coverage and independence

    Marketplace lists US$0.01 per detection call and US$0.13 per AI remediation call. The US$1 metering unit is not a per-call price. No transaction or invoice tested.

    View source
  • June Siteimprove MCP launch

    Publisher: Siteimprove

    Original language: English

    published date precision: day

    retrieved date precision: day

    Source publication date 2026-06-30

    Source retrieval date 2026-11-20

    Coverage and independence

    Prior MCP integrations already offered audit and remediation tools. Not evidence of the new purchase path.

    View source
  • August Siteimprove help

    Publisher: Siteimprove

    Original language: English

    retrieved date precision: day

    Source retrieval date 2026-11-20

    Coverage and independence

    The described direct MCP subscription route required Accessibility Pro or Advanced. Modified date, not necessarily first publication.

    View source

Gaps in the evidence

AWS independently hosts the commercial surface but does not independently test Siteimprove’s performance. No paid execution, invoice or user outcomes measured.

Explore the research

Research layers

Research entries

  1. Layer name

    01 Money flow

    Research and limitations

    The listing quotes US$0.01 per detection call and US$0.13 per AI remediation call. No invoice observed.

  2. Layer name

    02 Orders

    Research and limitations

    No verified order or deposit.

  3. Layer name

    03 Bottlenecks

    Research and limitations

    Review staff may be the constraint; no time measured.

  4. Layer name

    04 Beneficiaries

    Research and limitations

    Web teams and tools sellers may benefit; not demonstrated.

  5. Layer name

    05 Company finances

    Research and limitations

    No company financial effect verified.

  6. Layer name

    06 Costs

    Research and limitations

    Staff time illustration is not paid savings.

  7. Layer name

    07 Ownership

    Research and limitations

    No ownership or trading inference.

  8. Layer name

    08 Hiring

    Research and limitations

    No staffing change verified.

  9. Layer name

    09 Government

    Research and limitations

    No government budget claim.

  10. Layer name

    10 Language

    Research and limitations

    English sources; Denmark issuer and AWS USD billing.

  11. Layer name

    11 Behaviour

    Research and limitations

    No buyer use or retention verified.

  12. Layer name

    12 Contradiction

    Research and limitations

    June functionality and August subscription predate new billing.

  13. Layer name

    13 Missing evidence

    Research and limitations

    No account-specific quote, invoice, reader outcomes or measured fixes.

  14. Layer name

    14 Commercialisation

    Research and limitations

    Listing appears live; activation/payment not tested.

  15. Layer name

    15 Alternatives

    Research and limitations

    Compare existing manual and subscription coverage.

  16. Layer name

    16 Geography

    Research and limitations

    Check local availability and currency terms, not established here.

  17. Layer name

    17 History

    Research and limitations

    No prior verified OTR outcome claimed.

  18. Layer name

    18 Pricing

    Research and limitations

    Both listed rates exclude direct evidence of staff cost, tax and any other platform charges.

  19. Layer name

    19 Disproof

    Research and limitations

    Equivalent checks at lower full cost undermine the purchase case.

  20. Layer name

    20 Action clock

    Research and limitations

    Fresh from 20 Nov 06:06 SGT until 21 Nov 06:06 exclusive; stop if no cost or evidence advantage.

Back to Signals

AI powered research