Yeah, this is still one of those “boring but worth it” areas.
In real projects, the biggest wins I’m seeing in 2026 are usually not some fancy trick — it’s just removing junk and getting the page to render faster.
What’s worked best for me:
– **Fix LCP first**
– Compress the hero image hard
– Serve WebP/AVIF
– Don’t lazy-load the main above-the-fold image
– Preload the hero image if it’s the LCP element
– Cut down on sliders/video backgrounds if you can
– **Kill heavy JS**
– Delay non-critical scripts
– Defer stuff that doesn’t need to run right away
– Be ruthless with third-party tags: chat widgets, heatmaps, ad scripts, social embeds, etc.
– A lot of sites are slow just because they’re loaded with marketing junk
– **Fonts matter more than people think**
– Use fewer font families/weights
– Self-host if possible
– Preload the main font
– Use `font-display: swap`
– I’ve seen small font cleanup improve feel a lot, even if the lab scores only move a bit
– **CLS is usually theme/plugin nonsense**
– Set width/height on images and embeds
– Reserve space for ads and banners
– Avoid injecting cookie bars or popups that shove content down
– Watch for late-loading elements from page builders
– **WordPress cleanup helps a ton**
– Too many plugins = slow DOM + extra requests
– I usually remove anything that duplicates functionality
– Page builders can be okay, but some setups are bloated as hell
– Sometimes switching theme does more than 10 “optimization” plugins combined
A few practical things I’d check on every site:
1. **Run the page through PageSpeed + WebPageTest**
2. **Look at the actual LCP element**
3. **See what’s blocking render**
4. **Disable one plugin/script at a time and retest**
5. **Check mobile first, not desktop**
Also worth saying: don’t obsess over perfect scores if the site is already fast enough in real use. I’ve had sites with “meh” PSI scores still rank fine because the content and CTR were strong. But if you’ve got bad INP or CLS, that can definitely