PageRank AI (legal entity "PageRankAI") is a Shopify SEO/GEO optimization app whose docs are heavily tuned for AI/LLM consumption — it ships an llms.txt, an llms-full.txt, and a deliberately keyword-stuffed resource hub — yet the foundational pages they point machines and merchants to are broken, self-contradicting, or point at an install path nobody can find. Of the 13 "Primary pages" in llms.txt, 7 are reachable, 3 return HTTP 500, and 3 (/demo, /top-shopify-apps, /best-shopify-apps-small-business) were not fetched and remain unverified.
1. The same entry tier is priced 8x apart — and named two different things — on a single page (critical)
Problem: The pricing card reads "Limited Special — $19.69 /mo … Up to 20 product pages." The FAQ on the same page states: "The starter plan covers up to 20 product pages at $2.49/month" and again "The starter plan ($2.49/month for first 20 customers) covers up to 20 pages." Same entry tier, same "up to 20 pages" scope, two prices ($19.69 vs $2.49) that differ ~8x. The names don't match either: the card is titled "Limited Special," but the FAQ calls this tier "the starter plan" (and refers to the higher tiers as "the Prometheus plan" / "the Maximus plan" while the cards read "Frontier Prometheus" / "Frontier Maximus").
Consequence: A merchant — or an AI shopping/answer agent extracting the price to cite — cannot determine what the entry plan costs or what it is called. An agent indexing this page (the page that doubles as llms-full.txt) will surface whichever number and name it parses first, with no way to know they conflict. Either figure could end up quoted back to a customer who then disputes the charge.
The fix: Pick one entry price and one plan name, state each once, and align the FAQ to the card labels. If "$2.49 for first 20 customers" is a promo, mark it as an explicitly labeled, dated discount off the $19.69 list price.
2. Three of the 13 "Primary pages" in llms.txt return HTTP 500 — including the main explainer (critical)
Problem:llms.txt lists /what-is-pagerank-ai as its first content/explainer entry ("What is PageRank AI?", second only to the homepage), but it returns HTTP 500 (confirmed via WebFetch and TinyFish). So do /pagerank-ai-vs-semrush-shopify and , both also listed as Primary pages. The dead explainer is the destination of the "Learn more about what PageRank AI does" link on /resources and the "What is PageRank AI?" anchor on the GEO guide, the FAQ, and the comparison page. (Coverage note: of the 13 Primary pages, 7 fetched cleanly, these 3 return 500, and 3 — , , — were not checked.)
Code verification
Runtime snippet checks
Run the public examples from this audit in an isolated sandbox and attach the pass, fixed, skipped, and failed counts to the report.
Consequence: The most-linked explainer on the site — the one that defines the product — is dead, and the two pages a merchant would read to decide against a competitor are dead too. An LLM following llms.txt (the whole point of shipping one) hits 500s on top entries and may drop the site from its index or cite nothing. Human visitors clicking "Learn more about what PageRank AI does" land on an error.
The fix: Restore the three pages (a 500, not a 404, suggests a server/render fault, not just deletion). Until fixed, remove them from llms.txt and repoint the inbound links, so machines and humans aren't routed to errors.
3. Docs instruct installing a Shopify app whose App Store listing could not be located (critical)
Location: Homepage CTA "Get the app"; "How does it work? → Install App — One click to add our app to your website"; FAQ "Where can I start?"
Problem: The primary onboarding instruction across the site is to install the Shopify app: "Install the Shopify app, paste your highest-converting product URLs into the analyzer… from install to first optimization — takes under 10 minutes." The documented install path could not be verified: no apps.shopify.com/pagerank listing surfaced via WebSearch or TinyFish search (those searches returned competitors MagicGEO, StoreSEO, Smart SEO — but no PageRank.ai listing) as of 2026-06-26. This is absence-of-evidence, not proof the app doesn't exist — but the documented first step cannot be followed from the docs alone.
Consequence: The documented first step — "one click to add our app" — has no findable target, so a reader cannot complete it. Every downstream instruction ("paste your product URLs," "deploy the recommended changes") then depends on a step the docs don't actually enable. At minimum the install path is undocumented; at worst it's dead.
The fix: If the app is live, link the exact apps.shopify.com/<handle> URL from every "Get the app"/"Install App" CTA. If it is unlisted/in review, say so explicitly and provide the real access path (direct install link, waitlist, or "book a demo" as the actual entry point) instead of implying a one-click App Store install.
4. The GEO guide preaches one brand name while the site's own index ships seven (significant)
Location: /shopify-geo-optimization-guide vs. /llms.txt vs. /privacy + /terms
Problem: The GEO guide's "Align entity signals" section tells readers: "Use consistent brand naming ('PageRank.ai', not multiple variants)." Meanwhile llms.txt deliberately lists seven brand variants — "PageRank AI, Page Rank AI, pagerank ai, page rank ai, Pagerank AI, PageRank.ai, Rank AI (contextual)" — and the legal pages use yet another form, "PageRankAI" (one word). So the canonical name is simultaneously PageRank.ai, PageRankAI, and five other spellings.
Consequence: The product violates the exact entity-consistency advice it sells. For AI engines (the stated audience of llms.txt), inconsistent entity naming dilutes the trust signal the guide says is critical — by their own argument, the brand is harder to cite confidently. It also makes it impossible for an agent to pick the canonical name.
The fix: Choose one canonical brand string, use it in llms.txt, the legal entity references, and the UI, and demote the rest to a single "also known as" alias line. Practice the guide's own rule.
5. The scoring UI mixes "GRO" and "GEO," shows SEO on two scales, and the cards carry a second undefined tier-label set (significant)
Problem: The AI/citation metric is labeled "GRO Score — AI Citation Readiness" in the score header, "GEO Issues" in the issues panel, and "GEO --/800" in the comparison table — three labels (GRO/GEO) for one concept. Separately, "SEO Score" in the top cards is followed by "/800", while the SEO Analysis section and the Score Comparison both show SEO as "--/100" — the same SEO metric appears as both /800 and /100. This naming chaos extends to pricing: alongside the four card titles (Limited Special / Frontier Prometheus / Frontier Maximus / White Glove), the pricing block also carries a second, undefined label set — Gro, Win, and Enterprise interleaved between the cards — and the Maximus card's "Everything in GRO plan" references this undefined secondary layer rather than any titled card.
Consequence: A user can't tell whether their AI metric is "GRO" or "GEO," or what a "good" SEO number is when the denominator flips between 100 and 800 on the same screen. An agent scraping the score can't normalize it — 630/800 and 79/100 are indistinguishable as "the SEO score" without a fixed scale. And the dual tier labels mean "Everything in GRO plan" points at a label (Gro) that's never defined as a plan, so neither a human nor an agent can resolve what the Maximus tier actually inherits.
The fix: Pick one term (GRO or GEO) for the AI metric and use it everywhere; fix the SEO denominator to a single scale across the score card, analysis, and comparison; and either remove the Gro/Win/Enterprise secondary labels or define them and make "Everything in GRO plan" reference a named tier.
6. The "only … extends into GEO" claim is contradicted by the product's own comparison matrix (significant)
Location: /best-shopify-seo-app-2026 vs. /pagerank-vs-storeseo-semrush-storerank
Problem: The "best app" page states flatly: "PageRank AI is the only option that extends into GEO territory — making product pages citable by AI engines, not just indexable by Google." But the comparison matrix on the sibling page grades StoreRank.ai as "Yes" for "Designed for ChatGPT/Gemini/Perplexity citation workflows" — i.e., the docs themselves credit a named competitor with exactly the AI-citation/GEO capability the other page calls exclusive. The matrix even appends a hedge — "this is a directional framework based on public product positioning" — that undercuts its own definitive Yes/Partial/No grades of named third parties.
Consequence: Two docs pages give opposite answers to "is anyone else doing GEO?" A merchant comparing tools, or an AI answer engine asked "which Shopify apps do GEO," gets a flat contradiction from the same domain — and the unqualified "only" superlative is the kind of competitive claim that invites dispute when your own page grades a rival "Yes" on it.
The fix: Reconcile the two pages. Either soften "only" to a defensible differentiator ("the only one that scores both SEO and GEO in a single Shopify-native workflow") or update the matrix; don't grade a competitor "Yes" on the capability you call exclusive elsewhere.
7. Contact email differs between the pricing card and the legal pages (significant)
Problem: The White Glove pricing card says "Let's talk: support@pagerank.ai." Both the Privacy Policy and Terms of Service direct all contact — privacy questions and copyright/DMCA complaints — to "letstalk@pagerank.ai." Two different official addresses (support@ vs letstalk@) for the same company.
Consequence: A customer or an agent extracting a contact address has a coin-flip chance of emailing an inbox that may not be monitored. For the DMCA/legal path especially, mail sent to the wrong address can mean a missed legal notice.
The fix: Standardize on one contact address across pricing and legal pages, or clearly scope each (e.g., support@ for product, letstalk@ for sales/legal) and state which is which.
8. The standalone FAQ's "learn more" links all lead to 500s (significant)
Location: /shopify-seo-faq
Problem: The standalone FAQ ends its key answer with three inline links — "Learn more about PageRank AI | Compare with Semrush | Explore rank AI capabilities" — which point to /what-is-pagerank-ai, /pagerank-ai-vs-semrush-shopify, and /rank-ai-shopify-seo. All three return HTTP 500 (see Issue 2). The same dead trio is linked from the GEO guide's "Related" footer and the comparison page's "Related pages" footer. This second FAQ also overlaps the homepage FAQ but answers differently and omits pricing entirely, so it neither confirms nor resolves the $19.69/$2.49 conflict.
Consequence: Every "learn more" path out of the FAQ is a dead end, so a reader who wants depth after the "intentionally simple" answers hits errors. Because the FAQ states its phrasing is designed "so both humans and AI engines can extract facts cleanly," the broken outbound links defeat the page's stated purpose.
The fix: Repoint these links once the target pages are restored (Issue 2); until then, drop the dead links rather than ship a fact-extraction page whose every outbound reference 500s. Consolidate the two FAQs or cross-reference them so pricing isn't answered in one and absent in the other.
9. "Research & insights" is an empty two-line stub (minor)
Location: /research
Problem:llms.txt lists this page as "Research & insights," positioning it as a destination for evidence. The actual body is two lines — "Discover the latest in AI-powered commerce, optimization strategies, and the future of product discovery." — with no papers, no links, no data, no content of any kind. The page also took ~22s to respond.
Consequence: A reader (or an AI engine following the "Research & insights" label) arrives expecting material backing the product's GEO claims and finds an empty placeholder. For a tool whose entire pitch is credibility with AI answer engines, a content-free "research" page undercuts the authority signal it's trying to project — and the ~22s load time risks timeouts for crawlers.
The fix: Either populate the page with actual research/insights to match the label, or remove the /research entry from llms.txt until there's content. Investigate the slow response.
10. Legal terms make no pricing commitment despite specific dollar prices on the homepage (minor)
Location: /terms ("Fees, Trials, and Taxes") vs. homepage pricing cards
Problem: The homepage commits to exact prices and a trial: $19.69, $49.49, $99.99/mo, each "14-day free trial." The Terms of Service say only: "Some features may require payment now or in the future. If fees apply, we will present pricing or an order form before you incur charges." There is no SLA, no refund policy, and the Limitation of Liability caps recovery at "the amount you paid… in the 12 months before the event (or $100 if you paid nothing)."
Consequence: The binding legal document disclaims the very prices and free-trial promise the marketing page advertises, leaving no contractual basis for the trial terms or what a paying customer is owed. A developer or procurement reviewer evaluating the platform finds the commercial terms exist only in marketing copy, not the agreement they'd actually sign.
The fix: Reference the published pricing/trial terms from the ToS (or link an order form/DPA), and state at least a baseline support/refund expectation so the legal terms and the pricing page agree.
What they do well
They built for agents on purpose — shipping llms.txt, llms-full.txt, and an intentionally "simple phrasing so both humans and AI engines can extract facts cleanly" FAQ is the right instinct; the execution (dead Primary pages, conflicting facts) is what fails it.
The comparison content is concrete — named competitors, a capability matrix, and an explicit "directional framework" disclaimer show an attempt at honest positioning rather than vague superiority claims.
Legal pages are clear about AI data flow — the Privacy Policy plainly names that page content "may be sent to AI model providers (for example, Anthropic or OpenAI)," which is more transparency than many tools offer.
Top 3 recommendations
Fix the 500s first. Restore /what-is-pagerank-ai, /pagerank-ai-vs-semrush-shopify, and /rank-ai-shopify-seo (or pull them from llms.txt and repoint every inbound link). The most-referenced explainer on the site cannot be a dead page.
State one price, one plan name, one brand, one contact. Resolve $19.69 vs $2.49, the "starter" vs "Limited Special" name split, the GRO/GEO scoring labels and the Gro/Win/Enterprise secondary tier labels, the seven brand variants vs. the "use one name" advice, and support@ vs letstalk@. These are extractable facts AI engines will quote — every conflict becomes a wrong citation.
Make the install path real. Either link the actual Shopify App Store listing from "Get the app"/"Install App," or stop promising a one-click install and document the true access path. The documented first step must be completable from the docs alone.