Skip to content

What does automation miss? Manual and exploratory testing in 2026

Mykhailo PoliarushUpdated 6 min read

In 2026, manual and exploratory testing are no longer the default way to test but the layer that catches what scripts cannot see: accessibility barriers, confusing flows and failures nobody wrote a test for. The gap is large: in the WebAIM Million 2026 report, 95.9% of the top one million home pages had detectable WCAG failures, up from 94.8% a year earlier, and that count includes only the failures a tool can find.

Why do humans still test software in 2026?

Automation keeps growing, but it covers less than teams plan for. In TestRail's press release for its 2025 Software Testing and Quality Report, carried by Quality Magazine, teams automated 40% of their testing on average and aimed for 63% by the next year. The same release says most respondents still run functional, regression, end-to-end and smoke testing manually. TestRail sells test management tools, so treat its survey as one vendor's view.

AI does not remove that work either. In PractiTest's 2026 State of Testing report, a survey by a test management vendor, 44.1% of testers whose organizations do not use AI expect it to reduce reliance on manual testing, but only 30.9% of those who already use AI report that as a main benefit. Teams that use AI find it helps them test more complex scenarios, not skip human judgment.

The reason is simple. A script checks what someone already expected. A tester exploring the product finds what nobody expected, and the most expensive defects are usually in that second group.

What does accessibility data say about the limits of automated checks?

Accessibility is the clearest case. The WebAIM Million 2026 found an average of 56.1 detected errors per home page, 10.1% more than in 2025, and concluded that full WCAG 2 A/AA conformance was "certainly lower than 4.1%". Six error types account for 96% of everything detected.

Most common detected WCAG failures on the top 1,000,000 home pages

Low contrast text
83.9%
Missing alternative text for images
53.1%
Missing form input labels
51%
Empty links
46.3%
Empty buttons
30.6%
Missing document language
13.5%

Source: WebAIM, The WebAIM Million 2026 (automated WAVE scan of 1,000,000 home pages, February 2026)

Those numbers come from an automated scanner, and WebAIM itself warns that "absence of detected errors does not indicate that a page is accessible or conformant." The UK Government Digital Service guidance says automated testing can find only around 30% of likely accessibility problems and that services should also be tested with real users. Tool vendor Deque reports a higher share: across more than 2,000 audits, 57.38% of issues were found by its automated tests, yet those issues fell under only 16 of the 50 WCAG 2.1 AA success criteria. Either way, keyboard traps, confusing focus order, unclear error messages and screen reader flows still need a person.

Complexity makes it harder. In the WebAIM data, home pages using ARIA averaged 59.1 errors against 42 for pages without it, and page elements grew 14.3% in a single year. More markup means more ways to break assistive technology in ways a scan will not report.

There is also a legal deadline behind this. The European Accessibility Act applies to covered services provided to consumers after 28 June 2025, including e-commerce. Passing an automated scan is not the same as being accessible.

What usability problems only show up when a person uses the product?

Usability problems rarely fail a test, because the product works as built. It just works badly. Baymard Institute's cart abandonment research found that 17% of US online shoppers abandoned an order because the checkout was too long or complicated, and another 17% because the website had errors or crashed. The average US checkout shows 23.48 form elements by default, against an ideal of 12 to 14 that Baymard identified in usability testing.

The upside is large. Baymard estimates that the average large e-commerce site can gain a 35.26% increase in conversion rate through better checkout design, and its benchmark of 60 leading sites found an average of 39 potential checkout improvements per site. Baymard sells UX research, but its method is the point: people using the product, observed by researchers. That is what exploratory and usability testing do in a delivery team.

How do teams split manual and automated testing?

Most teams run both, and exploratory testing is often more mature than automation. In the PractiTest 2026 report, 61.5% of respondents rate their exploratory testing as structured (44.2%) or optimized (17.3%), compared with 52.5% for automation (38.2% and 14.3%). AI in testing is still early: only 14.5% of PractiTest respondents call it structured or optimized.

How mature are testing practices? Share of respondents by maturity level

How mature are testing practices? Share of respondents by maturity level
CriterionPracticeInitialExperimentingStructuredOptimized
Exploratory testing10.4%28.1%44.2%17.3%
Automation16.4%31.2%38.2%14.3%
Risk-based testing17.6%33.4%36.8%12.2%
AI in testing34.9%50.6%12.4%2.1%

Source: PractiTest, The 2026 State of Testing Report (vendor survey, thousands of respondents)

The split that works is by risk, not by habit. Stable, repeated checks belong in automated suites that run on every change. New features, complex user journeys, accessibility and anything that depends on human perception belong with skilled testers, and what they learn should feed the next round of automation.

What are the main challenges for manual and exploratory testing?

Four problems come up again and again in the data.

  • Measured on volume, not value. In PractiTest's survey, 56.4% of teams are evaluated on test coverage and 40.1% on automation coverage, while only 8.6% are evaluated on business impact. Exploratory findings do not show up in coverage metrics, so they are easy to underfund.
  • Pressure to ship fast. In TestRail's 2025 report summary, 58% of teams say rapid releases let defects slip into production, and the top challenges include end-to-end testing across integrated systems (33%) and involvement too late in the development cycle (32%).
  • No test strategy. Nearly 40% of individual contributors in the PractiTest 2026 report say test strategy is missing or underdeveloped, while only 25% of leaders agree.
  • AI aimed at output, not risk. In the same report, 70% use AI to create test cases but only 19.9% use it to identify risk. More generated tests do not tell a team where to look.

What does good manual and exploratory testing look like?

Good teams treat manual testing as a skilled, planned activity, not as clicking through a checklist.

  1. Test with charters. Each session has a mission, a time box and notes, so exploration is repeatable and its findings are reported like any other defect.
  2. Pick targets by risk. New features, payment and sign-up flows, integrations and recent incident areas get human attention first.
  3. Pair automated scans with manual accessibility checks. Run a scanner in CI, then test keyboard navigation, screen readers, zoom and error handling by hand, as the GDS guidance recommends.
  4. Watch real users. Short usability sessions on key journeys find the problems Baymard documents before customers do.
  5. Turn findings into tests. Every repeatable bug found by hand becomes an automated check, so testers spend their time on what is new.

Where should you start?

Start with one critical user journey. Run a time-boxed exploratory session, an accessibility pass with a keyboard and a screen reader, and a short usability check, then compare what you find with what your automated suite reports. The gap tells you how much risk you are carrying.

SDCLabs was founded in Krakow in 2011, and its founders created the open-source testing frameworks Codeception and CodeceptJS. Our senior QA engineers run charter-based exploratory sessions, usability testing and accessibility checks alongside automated suites, with AI assistance for test ideas and reports and people deciding what matters. Engagements start in 1 to 2 weeks. See our manual and exploratory testing services or contact us. For the automated side, see QA and test automation, and for AI across delivery, read AI in the SDLC: 2026 trends.

Sources checked on 2026-11-17.

Frequently asked questions

Is manual testing still needed when teams use test automation and AI?

Yes. In the vendor PractiTest 2026 State of Testing report, 44.1% of testers at organizations without AI expect it to reduce manual testing, but only 30.9% of AI users report that benefit. Exploratory testing finds the problems nobody wrote a script for.

Can automated tools check accessibility on their own?

No. UK GDS guidance says automated testing finds only around 30% of likely accessibility problems, and the WebAIM Million 2026 notes that no detected errors does not mean a page is accessible.

What are the most common accessibility failures on websites?

In the WebAIM Million 2026, low contrast text appeared on 83.9% of home pages, missing alt text on 53.1% and missing form labels on 51%. These and three other types make up 96% of detected errors.

What does exploratory testing find that scripted tests do not?

It finds unexpected failures and usability problems in real flows. Baymard found 17% of US shoppers abandoned an order over a complicated checkout and 17% because the site had errors or crashed.

Related posts

Get an estimate for your product.