What you can defer and what the platform locks down
Squarespace ships its own first-party JavaScript bundle that powers the editor, the live preview, the navigation, the carousel blocks, and other interactive elements. That bundle is loaded synchronously and cannot be deferred — there is no Squarespace setting to defer it, no Code Injection trick that overrides it, no plugin that disables it. What you control is the third-party scripts you have added: chat widgets (Intercom, Drift, Tidio), analytics beyond GA4 (Mixpanel, Heap, Hotjar), social embeds, marketing pixels, and any custom scripts pasted into Settings > Advanced > Code Injection. These are the scripts that benefit from async or defer.
Async vs defer — pick the right attribute
Async and defer both load the script in parallel with HTML parsing, so neither blocks first paint. The difference is when the script executes. async runs the script as soon as it finishes downloading, which can interrupt parsing. defer waits until parsing is done before running the script, in the order the scripts appear. For most third-party Squarespace add-ons — analytics, chat widgets, social embeds — defer is the safer default because the execution order is predictable. async is for scripts that have no dependencies and should run as soon as possible (Google Analytics is a common async case).
The third-party-script audit
Open Settings > Advanced > Code Injection. Look at the Header panel — every script tag pasted in there runs on every page. Make a list. For each script, decide: is it critical to first paint (almost never), does it need to run after DOM-ready (most analytics and widgets), can it defer entirely (true for most marketing pixels). Add async or defer to each tag. Save. Re-run PageSpeed Insights and check the INP improvement.
The rebuilt script pattern
After the audit, the Header Code Injection panel should look something like the block below: every third-party script tag carries async or defer, the order is predictable, and inline code is wrapped in DOMContentLoaded if it has any dependencies. The page-level INP score should drop within one PSI re-run for the Lab Data and within four weeks for the Field Data (CrUX rolling window).