naturespath.com · Full-site SEO audit · April 16, 2026
58/100
The GEO score measures how ready the website is to be found, quoted, and cited by AI-powered search tools: ChatGPT, Google AI Overviews, Perplexity, and Bing Copilot. These tools work differently from traditional search. Instead of returning a list of links, they read the content directly and synthesize an answer. To appear in those answers, content needs to be written, structured, and technically accessible in specific ways. This audit scores Nature’s Path across five dimensions that drive AI citation likelihood. The current composite is 45/100. The real gaps are Citability (content not written in quotable passages), Structural Readability (paragraphs not chunked for AI retrieval), and Multi-Modal depth — not access. Verified 2026-04-17: a single-request WebFetch pulled naturespath.com/robots.txt cleanly without Cloudflare intervention, confirming the site is physically accessible to normal-rate crawlers.
5 Dimension Scores
What this grid shows. The five weighted dimensions that compose the 45/100 GEO score, each rendered as a 0–100 card with its percentage weight and a color-coded status.
Per-dimension scoring reveals which levers move the composite hardest. Authority & Brand at 60/100 is already a strength because NP carries Wikipedia + Wikidata coverage. Citability at 28/100 and Structural Readability at 38/100 are the lowest scores, so content rewrites that produce quotable 134–167 word passages lift the composite faster than any other work. Technical Accessibility is passing.
How to read the colors. Green = passing (≥60 in this weighted model). Amber = needs work. Red = failing. Status labels call out the strongest (STRENGTH) and weakest (FAIL) dimensions to guide prioritization.
Citability (25%)
28
FAIL — 7.0 weighted
Structural Readability (20%)
38
FAIL — 7.6 weighted
Multi-Modal (15%)
42
NEEDS WORK — 6.3 weighted
Authority & Brand (20%)
60
STRENGTH — 12.0 weighted
Technical Accessibility (20%)
60
ACCESSIBLE — 12.0 weighted
Dimension 1: Citability — 28/100
What citability measures. AI engines quote passages, not pages. A citable passage is a self-contained 134–167 word chunk that answers one explicit question and names its entities in full. Paragraphs shorter than 134 words get skipped as incomplete; anything longer than 167 gets truncated or rewritten on the way into the answer. The Citability score is a count of how many passages on the site land inside that window.
Why the 134–167 word window drives NP’s 28/100. Almost no paragraph on naturespath.com lands in that window. The site’s most compelling claims (40 years organic, ROC pioneer, $57M Bite4Bite donation total, family-owned since 1985) are not written in any AI-extractable format today. Every page without one of those passages forfeits citation potential regardless of the domain’s authority, so the work is rewriting brand facts into passages that sit inside the window.
How to read the findings. Each failed list item below names a specific page or passage gap that a rewrite will close. The plain-language callout underneath explains the passage pattern itself so a copywriter can work the list without needing a GEO-scoring crash course.
What citability means (plain language)
When an AI is asked a question — “What is the most sustainable organic cereal brand?” or “Who makes Regenerative Organic Certified oatmeal?” — it scans web content for passages it can quote accurately. Citability measures whether your content is written in self-contained, quotable blocks that give the AI a clear, extractable answer it can attribute to you without misrepresenting what you said. The optimal passage length for AI citation is 134–167 words. Shorter than that, there isn’t enough context. Longer, the AI may pull only a fragment.
/pages/our-path has four content paragraphs totaling ∼164 words; longest is 47 words — none reach the 134-word citation threshold
/pages/about renders no unique page body content at all — HTML delivers navigation, not brand story
Blog post “Benefits of Plant-Based Protein” has proper H2 question structure (“How Much Protein Do You Need?”) but no author attribution, no datePublished schema, and content is 8 years old
No citations or source links on health claims — AI engines discount uncited statistics
Brand differentiation passages absent — no 134–167 word answers to “What makes NP different from Cascadian Farm?” or “What is ROC?”
NP’s most compelling claims — 40 years organic, ROC pioneer, family-owned, Bite4Bite $57M — are not written in any AI-extractable passage format on the site
Dimension 2: Structural Readability — 38/100
What this section shows. The heading hierarchy and chunking failures that compress NP’s Structural Readability score to 38/100. A plain-language definition of how AI parses content, followed by per-template findings across homepage, Our Path, About, collection pages, and recipes.
Structural readability is the scaffolding AI uses to retrieve answers. Question-format H2s like “How Much Protein Do You Need?” serve as anchors an AI can attach an answer to. Image-and-headline composites with no paragraph body are invisible to AI retrieval. The list below names the templates that need H2 / paragraph structure before any other GEO work compounds.
How to read the findings. Pass-marked items confirm patterns that already work (blog post H2 question format). Fail-marked items are rewrite tasks mapping 1:1 to Implementation Checklist items at the bottom of this page.
What structural readability means (plain language)
AI engines parse content in logical chunks. They look for clear heading hierarchy, short focused sections, direct answers at the start of each section, and predictable patterns they can follow (H2 question → paragraph answer → supporting detail). A wall of text, or a page whose only headings are product names and navigation labels, gives an AI nothing to anchor on.
Blog post H2s use question format (“How Much Protein Do You Need?”) — genuine citability advantage
Homepage: only two substantive H2s — “What Tastes Good?” (marketing prompt) and “Hungry For Savings?” (promotional). No brand, certifications, or “who we are” section
Our Path / About: render as image-and-headline composites (“Leaving The Earth Better”, “Family Owned”, “Always Organic”) — AI sees heading labels but no paragraph content beneath them
Collections: no editorial intro copy; cereal collection contains no text describing what makes NP cereal different
Recipes: structured format (Ingredients, Instructions) but no Recipe schema in JSON-LD, no H2 introduction explaining why the recipe uses an NP product
Dimension 3: Multi-Modal Content — 42/100
What this section shows. The schema, video, and social-media cross-reference signals that drive multi-modal AI understanding, followed by a check-list of what NP has versus what it lacks.
Multi-modal signals are how AI builds confidence in a brand. A page carrying Product JSON-LD, a YouTube VideoObject, descriptive alt text on certification badges, and a populated sameAs array is far easier for an AI to resolve to the NP entity than a page with images and no context. NP’s strong off-site media presence (YouTube, Instagram, Pinterest, TikTok, Threads) is a passing signal that isn’t wired into on-site schema.
How to read the findings. Pass-marked items confirm existing strengths. Fail-marked items are schema deployments and content fixes tracked in the Implementation Checklist. The Organization schema hygiene issues named here are the highest-priority Critical items on the list.
What multi-modal means (plain language)
AI engines, especially Google’s multimodal systems, form richer understanding of a brand when they can cross-reference text with images, videos, structured data (schema.org markup), and product feeds. A page with product images + descriptive alt text + a YouTube video explaining the brand story + JSON-LD markup is far easier for an AI to understand and trust than a page with images and no context.
Strong off-site media presence: YouTube channel, active Instagram, Facebook, Pinterest, TikTok, Threads, LinkedIn (YouTube presence correlates ∼0.74 with AI citation rates)
Product imagery is strong and visually rich
Every page on the site outputs only 2 schema blocks: Organization and WebSite. No BlogPosting, Recipe, Product, BreadcrumbList, NutritionInformation, FAQPage, AboutPage
Organization sameAs has 4 of 8 empty strings; @context is http://; name is HTML-encoded (Nature's Path)
No VideoObject schema despite YouTube content
Shopify product vendor field renders "NaturesPath" (no space, no apostrophe) — matches no canonical reference; Google cannot confidently resolve it to the NP entity
Dimension 4: Authority & Brand Signals — 60/100
What this section shows. The third-party authority signals AI engines weight when deciding whether a brand is real and citable — Wikipedia, Wikidata, Crunchbase, B Lab, Ahrefs DR. Followed by the known Wikidata property gaps that compress the score below 100.
Authority is where the GEO score is rescued. NP carries substantive Wikipedia coverage for the company and both founders, a complete Wikidata entity, and DR 71 backlink authority. AI engines treat this pattern as a confirmed real entity and weight NP content accordingly. The work below is cleanup (Wikidata property fixes), not new entity construction.
How to read the findings. Pass-marked items name the authority signals already present. The Wikidata gaps table names the specific properties to update. Each missing or wrong property is a short discrete edit on Wikidata tracked in the Implementation Checklist.
What authority signals mean (plain language)
AI engines don’t treat all websites equally. They weight content from brands that are mentioned in authoritative third-party sources — Wikipedia, Reddit, industry publications, Wikidata, and social platforms. A brand with a Wikipedia article, a Wikidata entity, YouTube presence, and mentions across Reddit threads is treated as a confirmed, real entity. A brand with none of those signals is essentially anonymous to an AI.
This is where the score is rescued. Nature’s Path has the strongest authority foundation of any dimension in this audit.
Wikipedia article at https://en.wikipedia.org/wiki/Nature%27s_Path — confirmed, substantive
Wikipedia articles for both founders: Arran Stephens and Ratana Stephens (strong credibility multiplier)
Wikidata entity Q6980861
Crunchbase profile, Google Knowledge Panel likely present (Freebase ID /m/0555vtf confirmed)
B Corp certification listed in B Lab public directory
DR 71 (Ahrefs) — meaningful backlink authority
Canada Organic Trade Association member page, UBC Library Research Guide, Encyclopedia.com entries
Known Wikidata gaps reducing the score
Property
Issue
Official website (P856)
Lists http://www.naturespath.com/ — wrong scheme, wrong form
Pinterest username (P3836)
Lists naturespath — should be naturespathorganic
Twitter username (P2002)
Duplicate entries
Headquarters (P159)
British Columbia (province) — should be Richmond BC (city)
YouTube channel ID (P2397)
Missing
LinkedIn company ID (P4264)
Missing
Employee count (P1128)
Missing
Subsidiary brands (P355)
Missing all 7 sub-brands
Google Knowledge Graph ID (P2671)
Missing
Dimension 5: Technical Accessibility — 60/100
What this section shows. The physical accessibility of the site to AI crawlers — robots.txt rules, server-side rendering behavior, HTTP response codes, and declarative signals like llms.txt and named AI-crawler stanzas. Followed by a per-crawler status table and the passing / failing signals that keep the dimension at 60/100.
Crawler access is the floor dimension. Once access is confirmed, additional points come from declarative welcome signals. NP’s robots.txt allows all major AI crawlers; the remaining 40 points come from adding declarative opt-in signals (llms.txt, named crawler stanzas, RSL 1.0 licensing) that tell AI engines how to use NP content.
How to read the findings. The crawler table confirms accessibility for the five major AI crawlers plus CCBot and anthropic-ai. The trailing check-list names the declarative signals still missing (llms.txt, named stanzas, RSL 1.0 licensing) that would lift the score toward 100.
What technical accessibility means (plain language)
Even if content is perfectly written and structured, AI crawlers need to physically reach it. “Technically accessible” means the crawlers can make a request, receive real HTML content, and read it without executing JavaScript. This dimension covers robots.txt rules, server-side vs. JavaScript-rendered content, response codes, and declarative signals (llms.txt, explicit AI-crawler directives) that document which crawlers are welcome.
Crawler
Purpose
robots.txt Status
Practical Status
GPTBot
ChatGPT index + training
Allowed (falls under *)
ACCESSIBLE
OAI-SearchBot
OpenAI real-time search
Allowed
ACCESSIBLE
ClaudeBot
Anthropic index
Allowed
ACCESSIBLE
PerplexityBot
Perplexity AI index
Allowed
ACCESSIBLE
Google-Extended
Google AI Overviews / Bard training
Allowed
ACCESSIBLE
CCBot
Common Crawl training data
Allowed
ACCESSIBLE
anthropic-ai
Anthropic training
Allowed
ACCESSIBLE
Access isn’t the bottleneck. What’s actually holding this dimension back from 100/100 is the absence of declarative AI-crawler signals and some volume-related accommodations. The site could do more to explicitly document welcome:
No llms.txt at https://naturespath.com/llms.txt (confirmed 404). Explicit AI-crawler summary document. See checklist.
No llms-full.txt companion file. Emerging convention where some AI platforms fetch an extended, full-content dump alongside llms.txt for deeper context. Publishing both is a one-time deploy and future-proofs against the convention standardizing.
No explicit AI-crawler stanzas in robots.txt — crawlers are allowed via the User-agent: * fallthrough but the site doesn’t document that intent explicitly. Adding named GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and CCBot stanzas makes welcome unambiguous.
No HTTP X-Robots-Tag response headers with AI-specific directives. robots.txt is the primary channel, but HTTP response headers add per-URL granularity and work even when a crawler skips /robots.txt. Configurable at the Cloudflare edge via Transform Rules.
No <link rel="license"> in the HTML head and no license field in any JSON-LD block. Without an explicit content-usage declaration, AI platforms default to conservative citation behavior. Link to a short /pages/ai-content-policy stating attribution and linking requirements.
No RSL 1.0 licensing declaration — no content licensing signal for AI training use (low priority; convention is still stabilizing, but worth adding alongside the llms.txt deploy).
Homepage HTML is 647 KB; About is 516 KB. AI crawlers that impose content-size limits (Perplexity truncates around 50 KB of extracted text) may skip the tail of the page. Tied to performance checklist’s DOM-bloat items.
Cloudflare AI Audit / AI Scrape dashboard toggle unverified. Cloudflare added a dedicated AI-bot control separate from the Verified Bots allowlist. Default behavior varies by plan tier; a “challenge” setting silently blocks even allow-listed crawlers. Dashboard → Security → Bots → AI Audit.
Sitemap <lastmod> coverage on Shopify-generated child sitemaps is inconsistent — spot-check shows some URLs missing lastmod entirely. AI crawlers use lastmod to prioritize revisits; missing values force a full re-crawl allocation and reduce the crawler’s appetite for incremental refreshes. Shopify platform-level issue, but worth surfacing to the NP engineering team.
hreflang coverage across NP’s 6 locales (en-US, en-CA, fr-CA, en-GB, en-MX, es-MX) needs automated verification. Correct hreflang declarations tell AI crawlers which locale answers which audience and prevent cross-locale citation confusion. Locale routing already exists via Shopify Markets; the signal needs to be verified at render time on every subject page.
Shopify SSR delivers base HTML before JS runs — base content is in static HTML, crawlers don’t need JS execution to read.
robots.txt clean: no AI-crawler disallows, no Nutch-style full blocks that would sweep up legitimate AI traffic. Sitemap declaration present and valid.
HTTP/2 multiplexing at the Cloudflare edge (HTTP/3 available on supporting networks). Parallel request streams benefit AI crawlers that fetch many URLs concurrently.
Conditional-GET support — Shopify and Cloudflare emit Last-Modified and ETag headers, so AI crawlers can skip unchanged content on revisit and preserve crawl budget for new and updated pages.
Platform-Specific Visibility Estimates
What this table shows. Per-platform visibility estimates (Google AI Overviews, ChatGPT, Perplexity, Bing Copilot) with the primary blocker stopping each surface from citing NP content at a higher rate.
Platform-specific blockers guide sequencing. The Organization schema hygiene fix unblocks Google AI Overviews and Bing Copilot at the same time. The llms.txt deploy improves ChatGPT and Perplexity together. A single schema or content deploy can lift two or three platforms simultaneously when the work maps against shared ranking signals.
How to read the blockers. Each primary blocker is the single highest-leverage fix for that platform. All four map to specific Implementation Checklist items at the bottom of this page. The closing callout reinforces that the entire dimension is additive content + schema work, no infrastructure required.
Platform
Estimated Visibility
Primary Blocker
Google AI Overviews
35/100
Organization schema malformed; no FAQPage / HowTo schema; thin brand content on About / Our Path
ChatGPT (GPT-4o web)
32/100
No llms.txt; no author schema on blog / recipe content; low citability passage density
Perplexity AI
35/100
No llms.txt; insufficient 134–167-word citable passages on brand pages; sub-brand entities not connected to parent
Bing Copilot
42/100
Organization schema incomplete; Bing Webmaster Tools sitemap not submitted; no Microsoft Merchant Center feed
The gap is content + schema, not access
Crawlers can reach the pages; the pages don’t yet contain the citable passages, entity signals, and structured data that AI systems reward when choosing what to quote. Every fix below is additive content or schema work — no infrastructure roadblocks stand in the way.
Tooling — Managing AI Search Visibility
Deep Blue will use Searchable and Scrunch as the operating stack for AI-search visibility work on Nature’s Path. The two platforms cover complementary halves. Searchable drives the AEO workflow (optimization + content + entity management). Scrunch drives AI-citation tracking and share-of-voice measurement across the major AI surfaces.
SearchableAEO WORKFLOW
Answer Engine Optimization platform — the workflow layer for on-site and off-site AI visibility work. Drives the optimization tasks below:
AEO content audits and passage-level rewrites for citability
Entity and brand-signal management across Wikipedia / Wikidata / knowledge graph surfaces
llms.txt generation + ongoing maintenance
Structured-data deployment and validation
Content freshness monitoring against the 60/90-day cadence
Semantic chunking / Q&A structure enforcement for AI retrieval
ScrunchCITATION TRACKING
Dedicated AI-search visibility tracker. Runs scheduled queries across ChatGPT, Gemini, Perplexity, Claude, and Microsoft Copilot, captures mentions and citations, reports movement over time:
Daily/weekly query runs against the branded + non-branded + competitor comparison query set
Citation presence + sentiment + factual accuracy tracking per platform
Share-of-voice monitoring vs. Cascadian Farm, Kashi, Purely Elizabeth, Catalina Crunch
Alerting on sudden visibility drops or factual errors in AI responses
Reports that feed directly into the monthly client update
How these tools map onto the Implementation Checklist
The checklist items below describe the work; Searchable and Scrunch describe the execution and measurement layer. AEO optimization tasks (schema hygiene, llms.txt, citable-passage rewrites, Organization entity completeness, content freshness enforcement) are driven through Searchable. Citation tracking tasks (baseline query set, per-platform visibility, competitor share-of-voice, monthly reporting) run through Scrunch. Tool identity is noted in-line on individual tasks where one platform is load-bearing.
LLM Visibility Reports
Snapshots of how Nature’s Path surfaces across ChatGPT and Perplexity. Each run queries the full 362-prompt research set (see llmrank/prompts/), detects brand and sub-brand mentions, and captures any naturespath.com citations. Click a run to expand its summary; the “Open full report” button on each row opens the full per-prompt per-platform breakdown in a lightbox.
Latest baseline · 2026-04-19
Weak brand visibility: 23.6% overall (171 of 724 queries mentioned Nature’s Path or a sub-brand) across chatgpt 20.7%, perplexity 26.5%. The 0.3% domain citation rate is the single clearest gap: LLMs mention the brand but almost never link to naturespath.com as a source. Closing that gap is what the GEO checklist items below are targeting — author attribution, citable-passage rewrites on /pages/our-path, cited-source hyperlinks, and FAQPage schema deployment.
Bright spots (NP is getting cited): Kids Cereal 65%, Fair Trade 50%, Competitor Comparisons 50%.
Critical gaps (0–10% visibility despite strong brand credentials): Philanthropy 0%, Extended Tracking 0%, Dietary Leadership 0%, Clean Label 0%. Each of these maps to specific checklist items below — the Bite4Bite program ($57M+ to food banks), Smart Bran fiber leadership, and clean-label credentials should be surfacing in these AI responses and aren’t.