List AccessLint-only checks and tighten wording in comparison doc

Author: kzhou314

AXE_VS_ACCESSLINT.md

@@ -1,6 +1,6 @@
 # axe vs. AccessLint
 
-The a11y scanner ships two built-in scan engines: [axe-core](https://github.com/dequelabs/axe-core) (the default) and [AccessLint](https://github.com/AccessLint/accesslint), a newer, lightweight ruleset. Both run against the live page, report WCAG violations, and share a large core of checks. axe is the mature, well-documented baseline; AccessLint adds extra WCAG 2.2 and best-practice checks. This doc explains what AccessLint adds and which combination to pick.
+The a11y scanner ships two built-in scan engines: [axe-core](https://github.com/dequelabs/axe-core) (the default) and [AccessLint](https://github.com/AccessLint/accesslint), a newer, lightweight ruleset. Both run against the live page and report WCAG violations, and the two mostly overlap. axe is the mature, well-documented baseline; AccessLint adds a small set of checks axe doesn't run by default. This doc covers what's different and which to enable.
 
 > 👉 Pick engines with the `scans` input. They run independently and file their findings as separate issues.
 
@@ -9,14 +9,14 @@ The a11y scanner ships two built-in scan engines: [axe-core](https://github.com/
 | `scans`                  | Runs            | Best when                                                                                          |
 | ------------------------ | --------------- | -------------------------------------------------------------------------------------------------- |
 | _(omitted)_ or `["axe"]` | axe only        | the mature default; lowest noise                                                                   |
-| `["accesslint"]`         | AccessLint only | you want its newer ruleset without axe's duplicates, but you give up axe's maturity and Deque docs |
+| `["accesslint"]`         | AccessLint only | you want AccessLint's extra checks without axe's duplicates, but you give up axe's maturity and Deque docs |
 | `["axe","accesslint"]`   | both            | maximum coverage, at the cost of duplicate findings                                                |
 
 ## At a glance
 
 |                   | axe-core                                                                                | AccessLint                                                           |
 | ----------------- | --------------------------------------------------------------------------------------- | -------------------------------------------------------------------- |
-| **Strength**      | Mature, widely adopted, well-documented baseline                                        | Lightweight second pass with newer WCAG 2.2 and best-practice checks |
+| **Strength**      | Mature, widely adopted, well-documented baseline                                        | Lightweight second pass; adds a few checks axe doesn't run by default |
 | **Coverage**      | WCAG 2.0/2.1/2.2 (A/AA/AAA) tags plus best practices; some 2.2 rules are off by default | WCAG 2.2 A/AA plus best-practice rules                               |
 | **Per-rule docs** | Deque University help URL on every finding                                              | Rule IDs + messages; less rich public per-rule docs                  |
 | **Main risk**     | Doesn't catch every WCAG issue automatically (~57%)                                     | Newer and less battle-tested; smaller ecosystem                      |
@@ -25,12 +25,21 @@ Both cover the same large core: alt text, link/button names, form labels, ARIA v
 
 ## What AccessLint adds
 
-AccessLint includes checks that may not be surfaced by the scanner's default axe configuration, including some WCAG 2.2 and best-practice checks. Findings link to the [AccessLint rules reference](https://github.com/AccessLint/accesslint/blob/main/core/README.md#rules-1). Examples:
+The engines share most of their rules (AccessLint ships ~93, axe ~104, and the bulk map to the same checks). Measured against the scanner's **default** axe run, these are the only AccessLint checks with no axe rule that fires. See the [AccessLint rules reference](https://github.com/AccessLint/accesslint/blob/main/core/README.md#rules-1) for the full catalog.
 
-- **Focus Visible (2.4.7)** — focusable elements without a visible focus indicator.
-- **Orientation (1.3.4)** — pages locked to a single screen orientation.
-- **Accessible Authentication (3.3.8)** — password fields that block password managers / paste.
-- **Audio transcript (1.2.1)** — `<audio>` without a text alternative.
-- Best-practice extras such as generic alt wording ("image", "photo") and paragraphs styled to look like headings.
+**No axe rule covers these:**
 
-axe has its own 2.1/2.2 and best-practice rules too (for example `avoid-inline-spacing` for text spacing), so expect overlap rather than a clean split.
+- **Visible focus indicator** (2.4.7 AA) — `keyboard-accessible/focus-visible`: focusable elements must show a visible focus indicator.
+- **Accessible authentication** (3.3.8 AA) — `input-assistance/accessible-authentication`: password fields must not block password managers / paste.
+- **Generic alt wording** — `text-alternatives/image-alt-words`: alt text shouldn't be "image", "photo", etc. (axe only flags alt that _duplicates adjacent text_).
+- **Presentational element with focusable children** — `aria/presentational-children-focusable` (adjacent to axe's `presentation-role-conflict`, but a separate check).
+
+**axe has the rule, but only as _experimental_, so the scanner's default run skips it:**
+
+- **Orientation lock** (1.3.4 AA) — `adaptable/orientation-lock` ↔ axe `css-orientation-lock`.
+- **Paragraph styled as a heading** — `navigable/p-as-heading` ↔ axe `p-as-heading`.
+- **Accessible name missing visible label text** (2.5.3) — `labels-and-names/label-content-mismatch` ↔ axe `label-content-name-mismatch`.
+- **Focusable element without a semantic role** — `keyboard-accessible/focus-order` ↔ axe `focus-order-semantics`.
+- **Large-table data cell without a header** — `adaptable/td-has-header` ↔ axe `td-has-header`.
+
+Everything else overlaps. Even where rule IDs differ the checks coincide — e.g. 1.4.12 text spacing (axe's single `avoid-inline-spacing` vs AccessLint's `letter-spacing` / `line-height` / `word-spacing`) and 1.2.1 audio (axe `audio-caption` vs AccessLint `audio-transcript`).