Measure page titles by rendered pixel width in Google search results, not by character count, and check them for truncation and title-link problems. Pass every title of a site in one call so boilerplate tails shared across titles are detected. Returns per title: px, full_width_units (px / 22), tier (safe/long/edge/cut), fails, warns; plus boilerplate_tails for the whole set and verdict 'fix' if any title fails. Model: Google search results on google.com with the zh-TW (Traditional Chinese, Taiwan) interface, Chrome on Windows, title font Arial 22px, desktop result column 652px, mobile first line 343px, measured 2026-09-18. Full-width characters (East Asian Width W or F: CJK ideographs, kana, Hangul, full-width punctuation, most emoji) count 22px; printable ASCII uses Arial's own advance widths at 22px; other half-width characters count 12.2px; zero-width characters count 0 and each grapheme cluster (such as an emoji ZWJ sequence) counts once. Kerning is not modelled. Tiers: safe <= 572px, long > 572px (warning), edge > 630px (must fix), cut > 648px (must fix). Japanese and Korean use the same full-width assumption but were not separately measured. Google may rewrite title links, so a passing title is not a guarantee of what is displayed. Fails: tier edge or cut; more than 2 separators (、 or ，) in a title over 572px; more than 1 full-width question mark (？); a trailing phrase after the last 、 or ， shared by more than 2 titles in the same call (boilerplate; a site name after a colon is not counted, as Google suggests that pattern); empty title; the same 2-4 ideograph run 3 or more times (keyword stuffing). Warnings: tier long; more than 2 separators in a title within 572px; text before the first ？：、，wider than the 343px mobile first line.

Read the first element from local .html/.htm files and run the same pixel-width checks as measure_titles across all of them. Each path may be a file (read whatever its extension) or a directory; directories are searched recursively for .html and .htm files, skipping node_modules and names that start with '.', and stop with an error after 20000 files. Only the title in the document head counts; titles in comments, , and similar are ignored. Reads only, never writes. Character references &amp; &lt; &gt; &quot; &apos; &nbsp; and numeric references are decoded before measuring. Model: Google search results on google.com with the zh-TW (Traditional Chinese, Taiwan) interface, Chrome on Windows, title font Arial 22px, desktop result column 652px, mobile first line 343px, measured 2026-09-18. Full-width characters (East Asian Width W or F: CJK ideographs, kana, Hangul, full-width punctuation, most emoji) count 22px; printable ASCII uses Arial's own advance widths at 22px; other half-width characters count 12.2px; zero-width characters count 0 and each grapheme cluster (such as an emoji ZWJ sequence) counts once. Kerning is not modelled. Tiers: safe <= 572px, long > 572px (warning), edge > 630px (must fix), cut > 648px (must fix). Japanese and Korean use the same full-width assumption but were not separately measured. Google may rewrite title links, so a passing title is not a guarantee of what is displayed. Fails: tier edge or cut; more than 2 separators (、 or ，) in a title over 572px; more than 1 full-width question mark (？); a trailing phrase after the last 、 or ， shared by more than 2 titles in the same call (boilerplate; a site name after a colon is not counted, as Google suggests that pattern); empty title; the same 2-4 ideograph run 3 or more times (keyword stuffing). Warnings: tier long; more than 2 separators in a title within 572px; text before the first ？：、，wider than the 343px mobile first line.