Methodology: how realserp measures the SERP

What we measure, and how

The tools measure the rendered pixel width of a title or description in your browser, using the canvas measureText method with the font set to Google's render stack. Google truncates on width, not character count, so the pixel reading is the one that matters. The full method, the render fonts, and the measured budgets are on the SERP pixel widths reference. One set of constants in the code feeds the simulator and both checkers, so a title that fits in one tool reports the same width in the others.

Honesty about precision

The mechanism is measured fact: Google cuts on rendered width, and we measure that width directly. The desktop budgets are well corroborated across the reference simulators and published studies. The mobile budgets and the character equivalents are softer, because Google wraps long text on phones and shifts rules by device, so we present them as conservative guides. Google also rewrites a large share of titles and many descriptions, so a well-fitted tag is necessary, not sufficient. We say so on the tools rather than implying a fit is a guarantee.

Sources

The 2026 pixel budgets were cross-checked against the constants that live tools actually ship: Portent, Search Wilderness, Screaming Frog, and SISTRIX, plus Zyppy, which measures how often Google rewrites titles rather than where it cuts them. Those sources genuinely disagree, and no one has published a fresh render-and-measure study of either device for 2024 through 2026. Where sources disagree we state the range on the SERP pixel widths page and ship the value the maintained tools converge on, and we run our own remeasurement pass so the numbers are current rather than inherited folklore.

Privacy is a technical invariant

Nothing you type, paste, or upload leaves the page. There is no server round-trip, no account, and no upload step. The favicon you add in the simulator is read locally with the browser's file reader and never transmitted. This is not only a promise; it is how the tools are built, and it is why they work on localhost and unpublished drafts.

We dogfood what the site teaches

realserp ships valid structured data on its own pages, a hand-written llms.txt that points AI models at the reference pages, and a robots.txt that allows the search and AI crawlers we want to cite us. The site is its own worked example.

No invented numbers

The pixel budgets are measured and sourced, not folklore, and they carry their method. We do not publish statistics we cannot source, and we do not use fabricated testimonials or results. When a number is an estimate, such as the AI Overview real-estate bar, we label it as one.