The B2B SaaS Website Audit Checklist: 6 Places You're Quietly Leaking Demos

A technical audit tells you what's broken. It doesn't tell you what it's costing you. Robots.txt, alt text, and sitemap XML can all be perfect, and your demo requests still won't move.

These are the six places we score when we run a Demo Leak Teardown, ranked by what actually costs a B2B SaaS site demos, not by what's easiest for a crawler to flag. The score that matters here is lost revenue, not a technical grade.

Work through all six and you've done manually what the Demo Leak Teardown above does automatically: real signals instead of a guess, and a ranked call on which ones actually cost you demos. Enter your URL at the top of this page and it runs the same six checks live.

1. Speed (Core Web Vitals)

A slow first load reads as broken before anyone reads a word of your pitch. For B2B, check desktop first. Your buyer is at a desk during business hours, not thumbing a phone; mobile traffic shows up (a LinkedIn click, a forwarded link) but it converts rarely.

The relationship between load time and bounce isn't a feeling, it's measured. Google's own research puts the bounce jump at 32% as load time goes from 1 to 3 seconds, 90% at 5 seconds, 106% at 6 seconds. Run any real volume of paid traffic through a slow page and that percentage is a line item, not a rounding error.

Self-check: Run your homepage and your money page, pricing or demo request, whichever is closest to a sale, through Google's PageSpeed Insights. Check the Desktop tab first, that's where your buyer actually is; glance at Mobile but weight it less. Look at the field data (real visitor data), not just the lab score, where it's available. Google's own thresholds: LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1. If your LCP is in the red, start there. It's the number visitors actually feel.

The fix: on most B2B SaaS sites the slowdown is one of three things: an unoptimized hero image or video, third-party scripts (chat widgets, trackers, embeds) loading before the page content, or a web font blocking render. Compress and lazy-load anything below the fold, defer non-critical scripts, and preload your primary font. None of this needs a redesign. It's a cleanup pass.

If your site comes back green across the board, good, that's one leak ruled out. Move to the next one.

How the Teardown checks it: real Core Web Vitals when we can pull them, otherwise time-to-first-byte and page weight as a proxy, and it only calls speed a leak when those numbers are actually bad, not by default.

2. Primary CTA and conversion clarity

A site with one clear next step converts. A site offering “Learn more,” “Get started,” “Talk to us,” and “Request a demo” on the same page converts nobody, because the visitor is doing the deciding, not you.

Self-check: Load your homepage and count the calls to action above the fold. Then open your three most-visited pages (check GA4 or your analytics for top pages). Most of the time those are articles or other informational pages, not pricing or product, so the right CTA there isn't "book a demo." It's an email-capture lead magnet you can nurture. But check first: if one of those top three is already a bottom-of-funnel page, pricing, product, demo request, that traffic is already close to a sale, and the direct ask is the right one there too. Reading each page cold, ask what happens if someone clicks the main button. If you can't answer in one sentence, neither can they.

The fix: pick one primary action per page, matched to where that page sits in the funnel. Book a call or Start your teardown on the bottom-of-funnel pages; a lead magnet CTA on top-of-funnel content. Word it as the specific next step ("Book a call," "Start your teardown," "Get the guide"), not a vague verb like "Learn more" or "Submit." Every secondary link should look and read as secondary, smaller, lower contrast, out of the way. A visitor who isn't sure what happens next doesn't click to find out. They leave.

How the Teardown checks it: it pulls every button and link label actually on your page (not the nav) and reads them against what kind of page it is: a booking page's job is booking, an article's job is capturing an email, so those get judged on different criteria than a pricing or services page.

3. Mobile, as a credibility signal

B2B buyers rarely close a deal from their phone, but they open your site on one constantly: a LinkedIn post, a cold email your SDR sent, a link a colleague forwarded in Slack. That first touch, on mobile, is a credibility check as much as your logo or your homepage headline. This isn't about lost mobile sales. It's about whether the site looks like it belongs to a company worth a second look.

Self-check: Open your site on your own phone, not a browser simulator. Load your homepage and whichever page people land on first (check your top landing pages). Look for: text you have to pinch to read, buttons you miss on the first tap, anything overlapping or cut off at the edge.

The fix: this is almost never a rebuild. It's usually a handful of fixed-width elements, a hardcoded desktop table, an oversized image, a nav bar that doesn't collapse, that need a responsive pass, not a new design.

How the Teardown checks it: whether your pages actually have a working mobile viewport tag. It won't guess at how cramped your layout looks on a phone from a desktop render, that's a judgment call that needs a real device, which is why mobile ranks lowest of the six unless it's badly broken.

4. SEO and schema/OG

SEO is the “found” half of “found and trusted.” Missing meta tags or schema doesn't just cost you search ranking. It also changes how your page shows up when someone shares your link in Slack or on LinkedIn, and a blank preview card is its own small credibility hit.

Two separate things live in this section: keyword placement, so search engines know what the page is about, and schema markup, so search and AI answer engines know what type of thing it is. Skip either one and you're invisible to demand you'd otherwise qualify for.

Self-check: View-source (or use a browser extension) on your homepage and one page deep. Confirm each has its own title tag and meta description, not the same one copied everywhere, and that the primary keyword you want that page to rank for shows up in three places: the title tag, the meta description, and the first or second sentence on the page. Then paste your homepage URL into Google's Rich Results Test and a social-preview checker (Meta's Sharing Debugger works). Confirm a title, description, and image actually render.

The fix: unique, specific title and meta description per page (not "Home | [Company]"), with your primary keyword in the title, the meta description, and the opening sentence. Add a correctly sized Open Graph image (1200x630) and schema markup, Organization at minimum, plus Article, FAQPage, Product, or Review, whichever matches the page. Schema tells Google and AI answer engines what kind of thing they're reading, not just what it says. Validate at zero errors on the Rich Results Test.

How the Teardown checks it: it reads your actual title tags, meta descriptions, keyword placement, and schema/OG markup off the page, not a generic crawler score.

5. AI visibility (AEO)

A growing share of your buyers are asking ChatGPT or Perplexity “who does X” before they type your name into Google. If AI answer engines can't cite you, you're invisible in a channel a competitor may already own.

Self-check: Open ChatGPT, Perplexity, or Claude and ask the question your buyer would actually type, tested two ways: by category ("best [category] for B2B SaaS") and by your actual industry or ICP ("best [category] for [your specific vertical]"). See if you come up for both, and if you do, check whether what it says about you is accurate.

The fix: AI answer engines lean on structured, quotable content: positioning that names your category and the industries or ICP you actually serve, not just what you offer; named case studies with numbers; pages that answer one question directly instead of burying it in paragraph three. If you're not showing up, that's usually a content-structure problem, not a technical one. Tighten your positioning to say who you're for, not only what you do, and make your strongest proof points easy to lift as a quote.

How the Teardown checks it: it reads your actual homepage and offer-page copy for specific, quotable positioning, including named industries or ICP, not just category, and proof. It doesn't run a live ChatGPT query; most B2B AI-visibility loss is a content-structure problem that shows up straight in your own page text anyway.

6. Trust and social proof

Every visitor is asking, whether they'd put it this way or not, “has this company done this for someone like me.” A generic “trusted by” logo strip with no names attached doesn't answer that. A specific one does.

Self-check: Scroll your own site like a stranger would. Count how many proof points name a real client, a real number, or a real named person, not just a star rating or an unattributed quote. Then check whether you have verified reviews on a third-party platform: Clutch, Capterra, G2, Product Hunt. Those carry more weight than a quote on your own page, precisely because you didn't write them. If your best case study is a paragraph of adjectives with no name and no number, that's the gap.

The fix: replace generic praise with specifics: named client, named result, named person giving the quote (name and title, not "Marketing Director"). Claim your profile on Clutch, Capterra, or Product Hunt and get a handful of verified reviews there; a third party vouching for you outweighs the same words on your own site. One strong, specific case study still outperforms five generic logos.

How the Teardown checks it: it reads your actual testimonials and case-study copy for a named client, a named number, or a named person, versus generic praise a stranger would skim past. It doesn't check Clutch, Capterra, or Product Hunt directly; claiming those profiles is still on you.

If you only fix one thing this week

Start with whichever chapter above made you wince. That's usually the real one.

Frequently Asked Questions

How long does this checklist take to run?

Doing all six yourself by hand takes longer than a coffee break, plan on a real sit-down, not a quick skim. If you'd rather skip the manual work, the Demo Leak Teardown runs the same six checks live in about two minutes and puts a number on what each leak is costing you. The fixes vary, some are a content edit you can make today, others need a developer.

Do I need developer access to run this?

No, for the self-checks. They're all external: PageSpeed Insights, view-source, your own phone, a chat window. Some of the fixes (compressing images, adding schema) need someone with access to your code or CMS.

What if my site scores well on all six?

Say so. If your checklist comes back mostly clean, you're already ahead of most B2B SaaS sites we see. Worth a lighter pass on whichever dimension scored weakest, rather than chasing a perfect score on all six.

How is this different from an automated audit tool?

An automated tool hands you forty flags with no judgment attached. This checklist, and the Demo Leak Teardown behind it, prioritizes which of the six actually costs you demos, not every metric a crawler happens to flag.

Want us to run this on your site instead?

You just did the self-serve version. The Demo Leak Teardown runs the same six checks automatically, live on your own site, and surfaces the 3 leaks that actually cost you demos: what each one's worth and the fix.

Show me my leaks

It runs live, right on the results page, a couple of minutes, no inbox wait.