Landing Page Optimization Guide: Audit by Traffic Source

Most landing page advice starts on the page: shorter form, bigger button, a new headline. That's the wrong end of the problem. A visitor arrives holding a promise, and the page either keeps it or doesn't.

Landing page optimization means changing the page so more visitors complete one conversion goal. The strategy here runs upstream first: list the promise each traffic source made, check the first screen against it, fix the break, and test only in ways that leave search visibility intact.

In this guide:

• Who writes the promise for each traffic source, including the two sources where Google writes it
• A per-source audit table and a rule for one URL that serves several sources
• A landing page optimization template with one illustrative filled example
• How Google grades the page behind a search ad
• Four rules for testing an indexed page without hurting rankings

Landing page optimization is changing a page so more visitors complete its single conversion goal, and it works when the first screen keeps the promise of the click. A published win is a hypothesis for your page, not a size, so check what a headline lift actually measures before you copy it.

Pick one conversion goal per page (a purchase, a sign-up, a demo request) and judge every change against it. A page with two goals gives you nothing to optimize against.

A dedicated landing page pays when a visit has one purpose that your regular site pages dilute. Here are the use cases where it earns its place:

• Paid campaign. The ad makes a specific offer and the page exists to complete it.
• Product launch. One product, one audience, one action, with no menu of other products.
• Email offer. The subscriber clicked a subject line; the page continues that sentence.
• Lead magnet. A guide, checklist or template traded for an address.

The same logic covers a product landing page: it is a landing page when it answers one source's promise and a catalog page when it doesn't.

The unit you optimize is the gap between each click's promise and the first screen. The promise has an author, and the author differs by source:

• Email and paid social: you write the subject line, the creative and the caption, so you control both ends.
• Search ads: Google assembles the ad. A responsive search ad takes up to 15 headlines and 4 descriptions for a single ad, per Google Ads Help, and the assets can be shown in any order.
• Organic results: Google Search Central says title link generation is completely automated and draws on page content and references to it on the web. Snippets are created from page content and can differ from one search to the next.

That last point changes the work. A responsive search ad is a pool of promises, so every headline in it must be one your first screen can keep in any combination. For an organic result, the promise is the title link and snippet Google shows for the query, which may not be the title you wrote. Read it from the results page.

Landing page optimization best practices only pay once the promise holds, and our post on CRO best practices sorts which page tactics carry evidence. If you wonder how to optimize a landing page first, start with this promise and leave the buttons for later.

The figure shows four stops: the source promise, the click, the first screen and the conversion action. Only the first screen is yours to change after the click, which is why the audit below starts there. Before you audit, a noise test for a low conversion rate shows whether the gap is real.

Each traffic source writes its promise differently before the click, so the audit lists who wrote it, where it appears and which first-screen element has to keep it.

If you only run one traffic source, you can skip the table and audit that row. Most pages receive several, and each source needs its own row.

Set a baseline by source: tag each source in analytics so results read per source. Without that, a blended rate hides the row that is failing. The table gives you one promise per traffic source, and each row gets its own check.

One URL, three promises is the common case: an ad, an email and an organic snippet all land on the same page. Here's the decision rule, in order:

1. Keep the strictest promise on the first screen. The strictest is the one naming a specific offer or number. If the email promises a free trial for small teams and the ad promises "project planning", the first screen leads with the trial.
2. Use dynamic text only for keyword-level variation of the same offer. Swapping "project planning" for "project tracking" in the H1 is a variation. Swapping a trial for a demo is a different offer.
3. Give a source its own page when its offer differs. Two offers on one first screen keep neither promise.

Then apply the last step to the pool itself: if the page cannot keep a headline in your responsive search ad, delete the headline. You change the promise instead of the page.

Copywriting craft for the headline itself belongs to our guide to landing page copy. Here the job is narrower: does the first screen repeat what the source said?

A landing page optimization template has five fields per source: promise wording, where it lives, first-screen element, mismatch, and test or ship.

Filling it in for one page shows which fix comes first.

Copy this layout, one row per source:

Here is one worked example, the kind of landing page optimization examples teams ask for. It is illustrative: a fictional project-management app called PlanBoard, one URL, three upstream promises (a search ad, an email and an organic snippet). It is not a client result and carries no conversion numbers.

The recall test shows why the first row fails. Show the first screen to someone for five seconds, hide it, and ask what the page offered. Then compare the answer with the exact words of each source. If the ad said "for agencies" and the person remembers "organized work", the promise broke before any button or form mattered.

The first fix is whichever row has no echo of the source's words. Page elements come after the promise holds, and each is a test candidate rather than a fixed rule. Use this landing page optimization checklist, six checks in all:

1. One primary action. Name the next step on the button ("Start free trial"), and keep one primary action per screen.
2. Navigation menu. Remove it if the visitor's intent is one offer; keep legally required links.
3. Field count. Ask only what the next step needs, and set it by visitor intent: a cold ad click tolerates fewer fields than a returning subscriber.
4. Proof next to the claim. Put the review or logo beside the specific claim it supports.
5. Mobile first screen. Check that the H1 and the action are both visible without scrolling on a phone.
6. Speed. Covered below, with Google's yardstick.

Behavior data tells you which kind of failure you have. This table reads the three signals:

For diagnosing a low rate across a whole site, start with the post on why your conversion rate is low. Where the landing page is the suspect, this table is the page-level step. Landing pages are also the first leak in the conversion funnel optimization sequence, and measured results from named programs are covered in our conversion rate optimization case studies.

Google grades landing page experience as above average, average or below average against other advertisers on the exact same search over 90 days.

That status diagnoses, and Ad Rank uses auction-time measurements.

Two separate facts sit here:

• The status is a report. It is one component of Quality Score, and Google Ads Help says Quality Score is not an input in the ad auction: it is a diagnostic tool.
• Ad Rank is a calculation. It incorporates your bid and auction-time quality, which can include auction-time measurements such as expected click-through rate, ad relevance and landing page experience.

The status is relative. Your page is compared with the pages of other advertisers on the exact same search, so a below average status means worse than the pages competing on that search. It does not mean your page has a fixed fault. The same page could read average on a query where rivals are weaker.

If you manage paid pages, that gives you a reading rule in three steps:

1. Read the status per keyword. Use the status as a pointer to the searches where rivals' pages beat yours.
2. Compare your first screen with the promise of that keyword's ad. Most gaps are promise gaps, covered in the audit above.
3. Judge the fix by cost per conversion. Higher quality ads can often lead to lower CPCs, per Google Ads Help, so a page fix can lower your click cost as well as raise your rate. Conversion rate alone hides the first effect.

Speed has its own yardstick. Google's web.dev guidance says Largest Contentful Paint should be 2.5 seconds or less, measured at the 75th percentile of page loads, segmented across mobile and desktop.

The percentile matters: it reads the slow end of your visits rather than the middle, so an average that looks fine can hide a slow tail. Our post on Core Web Vitals covers the fixes. For the wider channel comparison, see SEO vs PPC.

An SEO-safe test follows Google's four rules: no cloaking, rel canonical on variation URLs, a 302 redirect, and a test that ends when the result is clear.

Google Search Central's website testing guidance sets these four rules, and each has a reason:

1. Use a 302 redirect, not a 301. A 302 tells search engines the redirect is temporary, so they keep the original URL in their index instead of replacing it with the test page.
2. Put rel canonical on variation URLs, pointing at the original. You want search engines to group the test URLs as variations of the original, and Google prefers this to noindex because you don't want the original left out of the index.
3. Show Googlebot what users see. Showing one set of URLs to Googlebot and a different set to humans is cloaking, which breaks Google's spam policies whether you're testing or not.
4. End the test when it's done. An experiment that runs unnecessarily long may be read as an attempt to deceive search engines, and Google may act on it.

The figure's left panel shows the unsafe version, a response of "HTTP/1.1 301 Moved Permanently" with "Location: /landing-b". The right panel shows the safe one: "HTTP/1.1 302 Found" with the same location, plus a canonical link in the head of /landing-b that points back to the original URL.

These rules reach organic traffic for a reason tied to the audit. Google builds the organic title link and snippet from page content, so a variation that rewrites the first screen also rewrites the material the organic promise is built from. So a test changes the page and can also change what organic visitors were promised.

Write each hypothesis from an audit row, so a test starts from a recorded mismatch:

Because the [source] promised [wording] and the first screen shows [current element], changing [element] to echo the promise will move [metric].

For example: because the email promised a free trial and the first screen leads with a demo, making the trial button primary will raise trial starts from email visitors.

Choose the metric by source mix:

• Paid pages: cost per conversion, since page quality also moves click cost, as covered above.
• Organic pages: conversion rate, plus a check that the title link and snippet in the results still match the first screen.

How long to run a test depends on sample size, which our post on the minimum detectable effect works through. If traffic is thin, multi-armed bandit testing shifts traffic toward the better variant sooner, but the four rules above still apply to every indexed URL.

A landing page optimization strategy built on the click's promise ends with a next step you can take this week: fill in the template for your highest-traffic page, rewrite the first screen of its weakest row, and run the next test only after the promise holds.