Every week we run hundreds of SaaS landing pages through the Website Verdict scanner. The pages that climb fastest share a surprisingly small set of traits — and most of them have nothing to do with chasing more keywords. This is the full framework we use internally: the signals we score, the order we fix them in, and the failure patterns that keep otherwise-good products invisible in both Google and AI answer engines.
The five signals that move the needle
When we sort our scan history by which fixes correlated with ranking gains, a clear hierarchy emerges. These are the levers we pull first, in order of impact.
1. One unambiguous H1 that names product and outcome
The single strongest predictor of a page that ranks and gets cited is an H1 a machine can restate without guessing. 'Acme — Invoice reconciliation for Stripe merchants' beats 'Money movement, reimagined' in every scan cohort we have. Clever taglines are fine in the hero visual; the H1 itself has one job: identify the entity and the outcome. When an LLM summarizes your page, the H1 is the anchor it builds the summary around — make it a factual sentence, not a mood.
2. Structured data that lets engines cite you confidently
SoftwareApplication, Organization, and FAQPage JSON-LD are the minimum kit for a SaaS landing page. Structured data does two jobs at once: it earns rich results in classic search, and it gives answer engines typed facts — price, category, operating system, publisher — they can repeat without hallucinating. Pages in our dataset with valid SoftwareApplication + FAQPage schema get materially more AI citations than equivalents without it. See our AI readiness audit guide for the full schema checklist.
3. Entity clarity in plain language
Somewhere in the first screen of content, say what you are in words a parser cannot misread: who the product is for, what category it belongs to, what it replaces. AI engines resolve your site to an entity; if your copy never states the category ('an accounts-payable automation platform for mid-market finance teams'), the engine either guesses or skips you. The test we apply: could a stranger read only your hero section and correctly complete the sentence 'This is a ___ for ___'?
4. Core Web Vitals on the real hero, not a lab score
Most SaaS heroes fail LCP for a preventable reason: a full-bleed image or autoplaying video loaded without priority hints, plus a webfont swap that shifts the headline. Preload the LCP asset, set explicit dimensions on the hero media, and keep the H1 in HTML text rather than inside an image. Speed is a tiebreaker in Google rankings — but layout stability is also a readability signal for the crawlers that feed answer engines.
5. Internal links to proof
Landing pages that rank rarely stand alone. They sit at the center of a small cluster — pricing, docs, comparison pages, case studies — all linking back with descriptive anchors. Those links do three things: they pass authority, they give crawlers more entry points, and they give answer engines corroborating pages to check claims against. A landing page with zero inbound internal links is a page Google has been told is unimportant.
The scoring rubric we apply
| Signal | Weight | What a 10/10 looks like |
|---|---|---|
| H1 + entity clarity | 25% | Factual H1; category and audience stated in the first screen |
| Structured data | 20% | Valid SoftwareApplication, Organization, FAQPage JSON-LD matching visible text |
| Content answerability | 20% | Question-style headings, direct answers, FAQ section |
| Performance | 15% | LCP under 2.5s on mobile, no layout shift on the hero |
| Internal linking | 10% | 3+ descriptive inbound links from related pages |
| Trust signals | 10% | Real company info, security page, visible pricing |
Why AI search changes the playbook
Classic SEO rewards the page that best matches a query. AI answer engines reward the page they can summarize and attribute without hallucinating. That means crisp, factual copy and machine-readable structure now matter as much as backlinks. The practical difference: in classic SEO a vague page with strong links can still rank; in AI search a vague page is simply skipped, because the engine has a hundred clearer alternatives to cite. Our GEO vs SEO guide covers how the two systems overlap and where they diverge.
If an LLM can't confidently restate what you do in one sentence, it won't cite you — no matter how many keywords you stuffed in.
The failure patterns we see most
- The mystery hero — three screens of animation before the product is named. Engines index the first meaningful text; make it count.
- Schema-copy drift — JSON-LD claiming things the visible page never says. Google treats mismatches as spam signals; keep them identical.
- Keyword-splatter titles — five keywords piped together in the title tag. One primary phrase plus the brand outperforms it in CTR and clarity.
- Orphaned pages — landing pages launched from ads with no internal links, invisible to crawl discovery.
- JS-only content — key copy rendered client-side after interaction. Most AI crawlers read the initial HTML only; server-render what matters.
A worked example
A B2B scheduling tool came through the scanner scoring 54. The homepage H1 was 'Time, on your side.' The fix list: rewrite the H1 to name the category ('Meeting scheduling software for recruiting teams'), add SoftwareApplication and FAQPage schema, compress a 1.9MB hero image, and add internal links from four blog posts. Four weeks later the page had moved from position 40s to the top 20 on its primary term and picked up its first Perplexity citation — with no new backlinks. Foundation fixes compound; they make every future link worth more.
Run the same scan on your site
The framework above is exactly what the Website Verdict scan automates. Drop your URL into the cockpit on the home page and you'll get the same prioritized Fix Pack we use internally — and if you're an agency running this for clients, the white-label SEO reports put your logo on the output.