Go to any PageSpeed optimization agency's website. Scroll through their portfolio. What do you see? Screenshots. Cropped images of green scores. Neatly framed 98s and 100s floating on polished case study pages.
Now ask yourself a simple question: how do you know those numbers are real?
You don't. And that is the problem. The PageSpeed optimization industry has built its entire credibility model around the least verifiable form of evidence possible: static images. In an age where anyone with a browser can run a live test in 30 seconds, the reliance on screenshots is not just outdated — it is a deliberate choice. Understanding why agencies make that choice is the first step toward finding one that doesn't need to.
The Screenshot Problem
Screenshots became the default proof in web performance for a simple reason: they were easy. In the early days of PageSpeed optimization, before Google provided public APIs and before the testing infrastructure matured, capturing a screen was the fastest way to communicate results to a client.
The problem is that the industry never moved past this. Even as tools improved and live testing became trivially accessible, agencies kept relying on screenshots. The reason is straightforward: screenshots are controllable. A live test is not.
When you control the evidence, you control the narrative. You can test at 3 AM when traffic is minimal. You can use a cached result from a CDN edge node. You can run the test 50 times and pick the single best score. You can even — and this happens more than the industry would like to admit — simply edit the numbers in a graphics tool.
None of this is possible when you hand someone a live URL and say "test it yourself."
What Screenshots Can Hide
The vulnerability of screenshot-based proof goes beyond simple dishonesty. There are at least four distinct ways screenshots can misrepresent actual site performance, and most of them do not even require intentional deception.
1. Cache-Warmed Results
Google PageSpeed Insights returns two types of data: lab data (simulated) and field data (from real Chrome users via CrUX). A screenshot taken immediately after a CDN cache has been warmed — or after a recent field data collection window during low traffic — can show dramatically better numbers than the same site tested cold during peak hours. The screenshot captures a best-case moment, not a representative one.
2. Cherry-Picked Runs
PageSpeed scores fluctuate by 2-5 points between runs due to network simulation variance. A site that genuinely scores 88-93 might hit 96 on a lucky run. That single run becomes the screenshot. The client sees a 96 and assumes consistency. The reality is that the score will never be reliably above 93 because the underlying architecture has not been rebuilt — just tweaked.
3. Temporary Optimizations
Some agencies apply optimizations that look good in a test but degrade over time. Aggressive asset inlining that balloons HTML size. Cache headers that expire after a week. Third-party script blocking that breaks analytics or functionality. The screenshot is captured during the brief window where everything looks perfect, before the compromises become apparent.
4. Outright Fabrication
This is the uncomfortable truth the industry avoids. It takes approximately 90 seconds to open a PageSpeed screenshot in any image editor, change the score from 67 to 97, and export a pixel-perfect result. There is no watermark. No digital signature. No verification chain. A screenshot of a PageSpeed score carries exactly as much verifiable weight as a handwritten note claiming the same thing.
"If you cannot verify a claim in under 30 seconds, the claim has no value. In web performance, verification takes exactly one click — paste the URL into Google PageSpeed Insights and press Analyze."
The Verifiable Alternative
The alternative to screenshots is remarkably simple: provide the live URL and invite the client to test it themselves. Not a staging URL. Not a development environment. The actual production site, running with real traffic, real third-party scripts, and real content.
This is what we call the "live proof" model, and it changes the entire dynamic of the client-agency relationship. Instead of asking clients to trust a static image, you give them a button that links directly to pagespeed.web.dev with the URL pre-filled. They click it, Google runs the test, and they see the real score. No intermediary. No interpretation. No trust required.
The live proof model has one significant implication: you have to be good. There is no hiding behind a curated screenshot when anyone can test at any time. Your optimization has to work at 2 PM on a Tuesday and at 3 AM on a Saturday. It has to survive traffic spikes, content updates, and seasonal changes. The score has to be consistently high — not once, but every time someone runs the test.
This is, frankly, much harder than producing a good screenshot. Which is exactly why most agencies do not do it.
How to Spot a Screenshot-Only Agency
When evaluating a PageSpeed optimization provider, there are several red flags that indicate reliance on unverifiable evidence. These are not guarantees of poor quality, but they should prompt serious questions.
- No live URLs in their portfolio. If every case study shows a screenshot but no clickable link to verify, ask why.
- Scores shown only as images. If the proof is always a PNG or JPEG and never a live test link, the control is intentional.
- Resistance to independent testing. If you ask "Can I test the URL myself?" and get a vague answer, something is wrong.
- Vague methodology descriptions. "We optimize your site for speed" tells you nothing. Ask what specific techniques are applied, and whether they survive a content update.
- Single-score presentations. Showing only the overall score without LCP, FID, CLS, and TTFB breakdowns suggests the details do not hold up to scrutiny.
What Real Optimization Looks Like
Genuine performance optimization is not a single technique or a plugin installation. It is an architecture-first approach that addresses speed at every layer of the stack, from how the HTML is generated to how the final pixel is painted on screen.
Our approach involves a 30-technique playbook that covers font optimization, critical CSS extraction, image pipeline automation, resource hint strategies, JavaScript elimination, server-level compression, and much more. Each technique is measurable, each contributes to the overall score, and the combined effect is what produces consistent 95-100 results.
The key word is "consistent." A single test showing 100 means very little if the next ten tests show 82. What matters is the range. A well-optimized site should score within a tight 3-5 point band across dozens of consecutive tests, at any time of day, from any testing location. That kind of consistency only comes from architectural decisions — not surface-level tweaks.
This is why we publish every project with a "Verify Score" button that links directly to Google PageSpeed Insights. We do not ask you to trust us. We ask you to test us. Every project on our portfolio page is independently verifiable, right now, by anyone with a web browser.
That is the standard the industry should hold itself to. Screenshots served their purpose a decade ago. Today, there is no excuse for unverifiable claims. The tools are free. The tests take 30 seconds. The only reason an agency would avoid live verification is if they cannot survive it.
Frequently Asked Questions
See Verifiable Results Yourself
We publish every project with live proof. No screenshots, no trust required. Visit our portfolio, click "Verify Score," and see what real optimization looks like in Google's own testing infrastructure.
We are a performance engineering agency specializing in verifiable PageSpeed optimization. Every project we deliver is testable via Google PageSpeed Insights — no screenshots, no blind trust, just live proof.