Raw and encoded square brackets make different URLs, Google says

A URL with raw square brackets and the same URL using %5B and %5D instead are different URLs, a Google employee said on Bluesky on 19 October 2026. Search Engine Journal covered it on 24 October. The advice for sites that mix the two forms in links, canonical tags and sitemaps is to pick one.

What changed

Write a bracket as the character itself in one link and as %5B in another, and Google sees two addresses. The Google employee said they'd have assumed the two forms were equivalent, and had probably answered this wrongly for people before. Their post cited RFC 3986, the standard for URI syntax, which reserves square brackets and plus signs.

If a site serves the same content at both versions, the post said, Google crawls each one, folds the duplicates together and keeps one. Search Engine Journal adds that when links and canonical tags mix the forms, Google may end up choosing the version itself. The fix on offer was short: pick one form and use it everywhere, in internal links, canonical tags, the sitemap and structured data. A site that already uses raw brackets throughout should keep working, the employee said, though they wouldn't build a new one that way.

That lines up with Google's URL structure documentation, Search Engine Journal noted. The page says Google Search follows IETF STD 66 for URLs, and puts the rule in one line: "Characters defined by the standard as reserved must be percent encoded." It lists brackets as a way of adding parameters under not recommended, and recommends an equals sign between each key and value with an ampersand before each extra parameter. If a site's URLs miss the page's requirements, it warns, Google Search will likely crawl the site inefficiently, which can mean extremely high crawl rates or no crawling at all.

What it means for South African businesses

For South African sites, this means a check, and for most it'll be a quick one. Neither the Bluesky post, as reported, nor Google's URL structure documentation names South Africa. The URL rules are written for Google Search in general, so we'd treat them as applying here. If none of your page addresses contain square brackets, nothing changes. Where they do appear, we'd look first at filter and sort pages, the kind Google's URL documentation uses in its parameter examples. That puts ecommerce stores with product filters at the front of the queue.

Our view: leave working URLs alone and fix the mix. The employee expected sites that use raw brackets throughout to keep working, so we wouldn't rename addresses Google already crawls just to swap brackets for %5B. The trouble is a page linked one way in your menu and named the other way in its canonical tag or your sitemap. Google's canonical documentation already asks for consistency here: link to the canonical URL, and don't give one page different canonical URLs in different places. New builds are simpler. Keep brackets out of parameters from the start, which is one line in the brief. On a site that's already live, our technical SEO fixes include a crawl of every page, compared with Search Console.

What to do now

1. Open your XML sitemap in a browser, along with any sitemaps it lists, and press Ctrl+F to search for an opening square bracket, then for %5B. If neither turns up, your sitemap is clear. If only one does, note which form it is, so you can compare it with your links and canonical tags. If both appear, you've found the mix the Google employee described.
2. Pick one page whose address has brackets in it, open it in Chrome, press Ctrl+U to view the source and search for canonical. Compare that URL with its sitemap entry and with the links pointing to the page. Where they differ, ask your developer to bring the canonical tag, the sitemap, internal links and any structured data into line with the form your page addresses already use. Change the references and leave the addresses themselves alone.
3. Paste that page's address into the URL Inspection tool in Google Search Console. Google's troubleshooting guide says the tool shows which page Google considers canonical, so you can see which version it kept. Once the references match, Request Indexing asks Google to look again, but it's subject to quotas, so save it for your most important pages. Don't block either version in robots.txt or add noindex to force a choice. Google's canonical guide says robots.txt isn't for picking a canonical, since blocked URLs can still be indexed without their content, and that noindex blocks a page from Search completely. Our post on what robots.txt controls for AI crawlers covers what else that file can and can't do.
4. If a new site or a rebuild is on the go, ask your developer how filter and sort pages will build their addresses. Google's URL documentation recommends an equals sign between each key and value and an ampersand between parameters, with reserved characters percent encoded, so put that in the brief before any pages go live. On a rebuild, keep the addresses your current pages already use, or redirect each old address permanently to its new one. If you'd rather hand the whole check over, our SEO packages include a technical audit, and fixes for sites of up to 50, 300 or 1,000 pages depending on the package.

Sources

• Google Explains 503s, Robots.txt Limits & Bracket URLs: SEO Pulse, Search Engine Journal, 24 October 2026
• URL structure best practices for Google Search, Google Search Central, 21 October 2026
• How to specify a canonical URL with rel="canonical" and other methods, Google Search Central, 10 July 2026
• Fix canonicalization issues, Google Search Central, 21 August 2026