Why Are Recipe Websites So Slow? Page-Weight Data From 50 Sites

Published July 2026 · Updated September 2026 (clarifications only) · Data collected 20 July 2026

Quick answer: our July 2026 sample gives a picture of what recipe pages loaded, but does not establish why a particular site feels slow. Across 50 sampled recipe pages, the median transfer was 3.4 MB, with 120 network requests and 27 third-party hosts. On the 45 pages included in the ingredient-depth calculation, the median detected ingredients marker was 3,925 pixels down the page — about 4.4 screens at our 900-pixel browser height. These are observations from one visit per page, not speed scores.

What this study does not measure: loading speed, Core Web Vitals, crashes, memory use or ad refreshes. Third-party hosts are not necessarily ad or tracking services, and low page weight does not prove a site is ad-free. These July observations do not rank every recipe on a site or guarantee its current performance. The measurements have not been rerun.

What we measured, and how

We build Drizzlelemons, an ad-free recipe converter, so we have a commercial interest in this topic. We used the same browser procedure for each sampled page and report the sample and exclusions below. The sites came from our own conversion data, not a representative sample of the whole web.

• The sites: recipe domains selected from our own July 2026 conversion data — BBC Good Food, AllRecipes, NYT Cooking, RecipeTin Eats, Delish, Serious Eats, Sally's Baking Addiction, and others. One recipe page per domain; exclusions are listed below.
• The browser: headless Chromium at 1,280 × 900 with an en-GB locale, on a UK connection as recorded for the July run. Each page used a fresh browser context with no ad blocker and no cookie banners dismissed. This desktop browser run does not represent every mobile device, location or consent choice.
• The load: after the initial DOM loaded, the script waited three seconds, scrolled halfway down, waited 2.5 seconds, scrolled to the bottom, waited 2.5 seconds, then returned to the top. This could trigger lazy-loaded content; it did not wait for every possible background request or refresh.
• The numbers: transfer totals came from completed network requests and request counts from browser request events. Third-party hosts were counted using a hostname-suffix check against the final page host, not an ownership or advertising-service classification. External scripts means script elements with a src attribute; it does not measure how long scripts ran.
• Exclusions: the July results excluded navigation failures at thestonebakeovencompany.co.uk and jenneatsgoood.com, and bot-check responses at taste.com.au and thecountrycook.net. That left 50 publisher pages from 54 attempted. The separate Drizzlelemons example is excluded from those medians.
• Ingredient depth: the script looked for an ingredients heading or an element with an ingredient-related class or ID. The summary excluded two pages without a marker and three matches below 300 px, treated as jump-link matches, leaving 45 pages. This is a heuristic position measurement, not a manual check of the first readable ingredient on every page.

The headline numbers

Data units corrected 15 September 2026: the saved file labelled bytes divided by 1,024 as KB (technically KiB). We now show decimal MB and KB consistently. The median is the same recorded 3,303 KiB, approximately 3.4 MB; the earlier article labelled it 3.3 MB. This is a unit correction, not a new crawl.

The heaviest recipe pages we measured

The table reports one visit to each sampled URL. Page contents, caching and background activity can change the transferred amount on another visit. The total does not separate recipe text, photos, video, advertising or other resources.

How far down is the actual recipe?

On the 45 pages included in the ingredient-depth summary, the detected marker started a median of 3,925 px from the top — about 4.4 screens at the measured viewport height. On 20 of those 45 pages, it began more than 5,000 px down. The deepest match was 36,472 px on Alexandra Cooks' sourdough guide, about 41 screen heights. The script did not classify the content above those markers. A source page's jump link or the recipe URL trick can help readers reach the recipe without scrolling through the full page.

What about crashes, layout shifts and refreshing pages?

This study does not establish the cause of those problems. It counted requests, scripts, iframes and page positions; it did not measure memory use, crash rates, layout shifts, long tasks or ad refresh timers. Identifying the cause on a particular page would require a separate performance trace and testing on the affected device.

One converted recipe page for context

The separate Drizzlelemons page measured in the July run transferred 744 KB (727 KiB in the saved file), made 63 requests, contacted 1 third-party host, contained zero iframes, and had a detected ingredients marker 544 px from the top. Its transfer was about 22% of the publisher sample median. This was one converted recipe page measured with the same procedure, not a paired comparison of all 50 recipes or a current benchmark of Drizzlelemons. These counts do not establish loading speed or identify tracking activity.

What do third-party host counts tell us?

Third-party hosts can serve ads, images, fonts, analytics and other functions. A hostname count alone does not identify their purpose or their contribution to loading time. Likewise, page height does not tell us why a publisher chose its layout. For a separate discussion of the reading experience, see why online recipes are so annoying.

Use this data

You can cite the figures on this page with attribution and a link to the methodology. Describe them as a single July 2026 desktop crawl of selected recipe pages, and include the relevant sample size. For the saved per-site JSON and URL list, email contact@drizzlelemons.com. The September clarification checked the existing results against the measurement script; it did not rerun the crawl.

Frequently asked questions

Why are recipe websites so slow?

Our July sample recorded a median transfer of 3.4 MB across 120 requests, with 36 external script elements and 27 third-party hosts. These figures describe what loaded during the observation window. They do not measure loading speed, interaction delays or the share caused by advertising.

Why do recipe websites have so many ads?

Many recipe publishers use advertising to fund their work. Our measurements do not count ads or explain why each publisher chose a particular page layout; the tables report the page weight and positions we observed.

Why do recipe websites keep crashing?

This crawl did not measure crashes or identify their causes. Script and iframe counts alone cannot explain a crash on a particular device; that requires separate testing of the affected page and browser.

Why do recipe pages jump around or keep refreshing?

The crawl did not record layout shifts or refresh timers, so it cannot attribute page movement to ads or any other cause. It observed each page for eight seconds after the initial DOM loaded, rather than monitoring a whole cooking session.

How were these measurements taken?

Headless Chromium at 1,280 × 900 with no ad blocker, using one selected recipe page per domain and an eight-second wait-and-scroll sequence after the initial DOM loaded. The July results retained 50 publisher pages after four exclusions; ingredient-depth statistics used 45. Transfer and request figures came from browser network events on 20 July 2026.

Can I cite these numbers?

Yes — the data and methodology are free to use with attribution and a link to this study. For the raw per-site JSON, email contact@drizzlelemons.com.

Want a clean recipe view? Try an accessible recipe URL in the recipe converter. There is one free anonymous conversion per day; if the source blocks extraction, paste recipe text you can access.