Supply chain execution is bought on a long, committee-driven cycle that now starts with someone asking an assistant which vendors run warehouse, transportation and order management on one platform — and in a category where the shortlist forms before a salesperson is ever contacted, the vendors that establish citation visibility now lock in a structural advantage while the rest of the market is still treating AI search as a curiosity. Before we run the audit, we need to make sure we're asking the right questions about the right competitors to the right buyers. This document presents what we've learned about Infios's market — your job is to tell us what we got right, what we got wrong, and what we missed.
Before we measure how often AI assistants name Infios in warehouse, transportation and order management conversations, these three signals tell us whether those assistants can reach infios.com, trust how current it is, and discover what's on it. All three are derived mechanically from the August 13, 2026 analysis of 49 pages.
Disallow patterns in the User-agent: * group of robots.txt match 11 sitemap URLs by accident, including the WMS Buyer's Guide — which matches /*-s- only because the apostrophe in "buyer's" slugifies to -s- — and two Gartner analyst-report landing pages. The other two high items are sitemap lastmod values stamped on republish rather than real content change, and an English blog corpus where the visible date on nearly every post is more than a year old.Allow: / and no restrictions. The problem is the catch-all group everything else falls into: Googlebot, Google-Extended, ChatGPT-User and Bytespider inherit ten wildcard Disallow patterns that block 11 sitemap URLs. Discovery has three further gaps: the homepage appears in no sitemap, a retired -old duplicate page is advertised in sitemap-en.xml with a recent lastmod, and llms.txt is served with every non-ASCII character double-encoded.Supply chain execution software is bought by a committee, over months, against a shortlist that forms early — and that shortlist increasingly forms inside an AI assistant before anyone fills in a contact form. 95% of winning vendors were already on the buyer's Day One shortlist across nearly 4,000 B2B purchase decisions averaging $300K–$400K each (6sense, November 2025), which makes the question of who an assistant names when a distribution executive describes their problem a commercial question rather than a marketing one. Infios sits in an unusual position going into that measurement: a platform with real scale and a decade of accumulated evidence, most of which is filed on the open web under names the company no longer uses. That is a first-mover opportunity and a liability at the same time, and it is the reason this engagement starts with an entity question rather than a content one.
This Foundation Review presents the inputs that determine what the audit actually measures. The competitive set decides which vendors Infios is tested head-to-head against and which are tested only for category awareness. The buyer personas decide whose search intent the queries model — a CIO, a distribution center director and a 3PL chief operating officer ask genuinely different questions about the same platform. The feature and pain point taxonomies supply the language those queries are written in, drawn from how buyers describe the problem rather than how the category markets to them. Underneath all of it sits the Layer 1 technical baseline: whether AI crawlers can reach infios.com, whether they can tell what is current, and whether what they retrieve resolves to Infios at all. This is what we're validating together before the audit runs.
The validation call is a working session with real consequences, not a walkthrough of a document. Two kinds of decisions get made there. First, input validation: are the right buyers and the right competitors in the right tiers, and is our outside-in read of Infios's capabilities honest? Every correction redirects query budget — one of the questions on the table would change the entire competitive set and the vocabulary every query is written in. Second, engineering triage: the Layer 1 findings are independent of everything we decide about the query set, and the highest-severity item is under a day of work, so your engineering team can start the day this document lands. The specific items on both sides are collected in the Pre-Call Checklist near the end.
User-agent: * group (e.g. /*-typ$), drop the bare /*-s- and /*-d- rules that are too short to be safe on any multi-word slug, and rename the buyer's-guide slug to wms-buyers-guide.lastmod and BlogPosting dateModified from the real last-edited timestamp, and fix the llms.txt double-encoding — neither depends on a decision from the validation call, both are engineering-owned, and together they are the difference between a freshness signal crawlers can use and one they discount across all 936 English URLs.Three things to know before you read the rest.
Purpose This is the input layer for Infios's GEO visibility audit. Everything below — the competitors, the buyers, the capabilities, the buyer frustrations — becomes the raw material for the buyer queries we run against AI assistants. If our read of the supply chain execution market is off, the audit measures the wrong thing precisely. That's why this document exists before the audit rather than after it.
Your Job Read for what's wrong, not for what's right. Every section ends with a specific question in purple — those are the places where our confidence is lowest or the downstream consequence of being wrong is highest. You don't need to prepare a written response; the Pre-Call Checklist at the end collects every question in one place so you can walk into the call with answers.
Confidence Badges Every entity carries a confidence badge showing how directly it was sourced. High means it came from a first-party source — infios.com itself, or consistent, specific language across review platforms. Medium means it was inferred from category listings, analyst coverage or competitor-authored content. One thing worth knowing about this engagement specifically: Infios is the March 2025 rebrand of Körber Supply Chain Software, the WMS carries HighJump and K.Motion heritage, and the TMS is the acquired MercuryGate platform — so nearly all the review-site and analyst evidence that grounds the "high confidence" ratings below is filed under names the company no longer markets. That's why the name-variant list in the next section is deliberately long, and it's why medium-confidence items are where your correction is worth the most.
The company profile sets the entity we track across every AI response — the name variants below are literally how we detect an Infios mention in an answer that never links to you, and for a company three brand names deep that detection is doing more work here than in a typical engagement.
→ Infios sells an SMB WMS edition alongside its enterprise and 3PL editions, and we've classified the company enterprise on the strength of 5,000+ customers across 70 countries — but headcount estimates range from roughly 1,400 to 2,300 depending on the source, which straddles the mid-market line. Which segment do you want this audit to represent? The answer changes the entire competitive set and the vocabulary every query is written in: enterprise buyers compare you to Manhattan Associates, Blue Yonder and SAP and ask about global rollouts and analyst placement, while SMB and small-3PL buyers compare you to Deposco, Extensiv and Made4net and ask about published pricing and six-week go-lives. The two sets barely overlap, and testing a blended set would measure neither well. A second, related check: we intend to count "Körber," "HighJump" and "MercuryGate" mentions in an AI answer as first-class Infios mentions rather than as separate vendors, because that's where a decade of analyst and review evidence still sits — confirm that's how you want it scored, because the alternative materially changes your measured visibility. And one thing we could not confirm from public sources: Manhattan, Blue Yonder, SAP, Oracle, Infor and Made4net all have confirmed 2026 WMS Magic Quadrant placements, but we could not verify Infios's own — analyst standing is a frequent tiebreaker in AI-generated vendor shortlists, so it's worth telling us where you landed.
6 personas: 3 decision-makers, 1 evaluator, 2 influencers — each modelled as a distinct search intent, because the queries the audit runs are written in their language, not Infios's.
Critical Review Area Personas are the single highest-leverage thing you can correct in this document. Each one drives a distinct cluster of queries; a persona that doesn't exist in your deals burns query budget, and a persona we've missed is a set of buyer questions the audit will never ask. Read these six for who is duplicated and who is missing before you read them for accuracy.
Data Sourcing Note Name, role, department, seniority, influence level, veto power and technical level come from the knowledge graph. Three of the six are grounded in observed evidence: the CIO and the Director of DC Operations come from review mining, and the 3PL COO from Infios's own multi-client and 3PL content. The other three are inferences — the SVP of Supply Chain Operations and the VP of Transportation & Logistics at medium confidence, and the Director of Omnichannel Fulfillment at low confidence, derived from the Order Management and store fulfillment product lines rather than from anyone observed in review data. The role descriptions, buying jobs and query focus areas are synthesized by us from those attributes plus the pain points each persona is linked to — correct them freely, they're our interpretation, not your data.
→ Does the CIO lead the WMS selection, or arrive as an architecture gate after supply chain has already picked? If she leads, the ERP-adjacent queries — SAP EWM, Oracle Fusion WMS, "keep it in S/4HANA versus buy best-of-breed" — become their own cluster and Infios is tested against ERP defaults; if she gates, those questions fold into a smaller integration set and the head-to-head budget goes to the pure-play WMS vendors instead.
→ What share of your pipeline is 3PL versus shipper — retailers, manufacturers and distributors running their own DCs? The 3PL buyer searches in a completely separate vocabulary (multi-client billing, client onboarding speed, per-client margin) and meets a different competitive set (Extensiv, Deposco, Made4net) than the shipper buyer does. If 3PL is a third of revenue we split the query budget proportionally; if it's the majority, the whole audit reorients around it.
→ Does the SVP of Supply Chain genuinely hold veto power, or budget influence with the signature sitting elsewhere — with the CIO, the CFO, or a steering committee? If it's influence only, we demote him to evaluator and the validation-stage queries (procurement, contract terms, vendor viability) move to whoever actually signs, which changes roughly a sixth of the query set.
→ Can the DC director's usability objection actually stop a selection, or is she consulted after the shortlist is already set? If she can stop it, the weak usability and reporting ratings become a defensive query cluster we test deliberately — "is Infios hard to use", "Infios learning curve" — because that's where competitors will aim AI-generated comparisons; if she's consulted late, those queries carry far less weight and the budget goes to the executive evaluation instead.
→ Is the TMS bought on the same cycle as the WMS, or as a separate decision on its own budget line? If it's separate, we need a standalone transportation query cluster tested against E2open, Descartes and Oracle Transportation Management — where "MercuryGate alternatives" is a live search term — rather than folding transportation into platform-level queries where Manhattan and Blue Yonder are the comparison.
→ Does a dedicated omnichannel fulfillment owner appear in your deals, or does order management get bought by the SVP of Supply Chain as part of the platform? This is the lowest-confidence persona in the set and it owns an entire query cluster on its own — the OMS and delivery-promise questions. If the role doesn't exist as a distinct searcher, those queries should be rewritten in the SVP's executive vocabulary rather than a fulfillment director's operational one, and they'd read very differently.
Missing Personas? These roles sometimes appear in supply chain execution deals — do they show up in yours? ERP program lead / Director of Enterprise Applications — in an S/4HANA or Oracle Fusion migration, the person running that program often decides whether warehouse execution stays native or goes best-of-breed, and they search in ERP language ("SAP EWM versus third-party WMS") that nobody else in this set uses. Director of Warehouse Engineering / Automation — where conveyors, sorters, AS/RS and AMRs are already installed, the automation owner evaluates orchestration and WCS fit independently of the WMS decision, and asks vendor-specific integration questions our current personas wouldn't. VP of Strategic Sourcing / Procurement — enterprise deals of this size usually route through a formal sourcing function that runs the RFP, and if they shape the shortlist they search for evaluation criteria and vendor viability rather than capability. Who else shows up in your deals?
6 primary + 5 secondary competitors identified — primary means direct head-to-head testing in the audit, secondary means category-awareness testing only.
Why Tiers Matter Tier assignments decide where roughly 36 to 48 of the audit's head-to-head queries land — phrasings like "Infios alternatives", "Manhattan Associates vs Infios", "best WMS for third-party logistics", "MercuryGate alternatives" and "unified WMS TMS OMS platform". Six primary competitors at six to eight queries each consumes that budget entirely; secondary vendors get tested only for whether Infios surfaces alongside them in category questions. Softeon is the one primary we're least certain of — it carries medium confidence and was tiered from category listings rather than from observed head-to-heads. If Softeon rarely reaches your shortlists, moving it to secondary frees five to eight queries for Made4net, which overlaps Infios almost module-for-module, holds 2026 Magic Quadrant recognition, and is currently tiered secondary on the strength of a smaller installed base alone.
→ Three things to settle here. (1) Who's missing? Name any vendor that shows up in your last ten competitive deals and isn't on this list — a vendor absent from the graph is a vendor the audit will never test you against. Every name here was assembled from category listings and analyst grids rather than from your CRM, so this is the section most likely to have a gap. (2) Is Softeon really primary, and should Made4net be? Softeon is the only medium-confidence primary and was tiered from mid-market WMS round-ups rather than observed losses — does it reach your shortlists, or is Infor who you meet in mid-market bake-offs? Meanwhile Made4net overlaps you almost module-for-module with a six-week Fast-Track implementation and 2026 Magic Quadrant recognition, and sits secondary. Swapping those two moves five to eight head-to-head queries. (3) Is anyone here wrong to be on the list? E2open and Descartes overlap Infios only on the transportation and global-trade flank, with no warehouse execution answer — if they never appear in a platform evaluation, testing them as competitors spends category-awareness budget on a comparison your buyers don't make.
12 buyer-level capabilities mapped — 6 rated strong, 3 moderate, 3 weak — and the buyer language under each is how the audit's capability queries will actually be phrased.
Run receiving, putaway, replenishment, wave and waveless picking, packing, and shipping across multiple sites with real-time inventory accuracy down to lot and serial
Run many customers in one warehouse with separate inventory, rules, SLAs, and activity-based billing — and onboard a new client without a services project
Plan, tender, rate-shop, and track freight across parcel, LTL, truckload, and ocean, with freight audit and payment and claims handled in the same system
Orchestrate conveyors, sorters, AS/RS, and autonomous mobile robots alongside human pickers without a custom middleware layer for every vendor
Set engineered standards, see who is actually productive, and get pickers hands-free and heads-up with voice instead of scanning screens
Change our workflows to match how we actually operate, without waiting for the vendor's next release or forking off the upgrade path
Show one accurate inventory position across DCs and stores, promise a realistic delivery date at checkout, and route each order to the cheapest node that can hit it
Have the system handle carrier emails, exception triage, and reroutes on its own inside the rules we set, instead of handing us another dashboard to watch
Connect cleanly to our ERP, our carriers, and our trading partners' EDI without a six-figure integration project every time we add one
Screens a warehouse supervisor can learn in a shift, not a system that needs three weeks of training before anyone is useful on it
Live in a handful of weeks with a defined scope and a fixed price, not a four-to-eight-month program that needs specialist consultants throughout
Build the report I need myself this afternoon, instead of exporting to a separate BI tool or opening a ticket for every new view
Which Three Carry the Differentiation Queries? Six of the twelve capabilities are rated Strong. The audit tests all twelve, but competitive differentiation queries will emphasize three — and picking them is a commercial judgment we can't make from the outside:
Which three of these best represent where Infios actually wins deals — not which demo best, which close?
→ Three ratings questions, and the first one is uncomfortable on purpose. (1) The three weak ratings are load-bearing. User Experience & Ease of Use, Implementation Speed & Time to Value and Reporting, Analytics & Self-Service BI are rated weak on consistent, specific review language across Capterra, SoftwareConnect and SelectHub — "NOT USER FRIENDLY," "stepping back 20 years," four-to-eight-month implementations, limited built-in BI. These are exactly where Softeon (configuration-not-code), Made4net (six-week Fast-Track) and Deposco (modern UX) will win AI-generated recommendations, so the audit needs them rated honestly to test the right defensive queries. Correct them with evidence — a published implementation benchmark, a UX release — rather than with preference, because an inflated rating here means we never test the flank you're actually losing on. (2) AI & Agentic Execution is rated moderate purely on recency. Infios Archer was announced 8 July 2026 and has no independent review or analyst coverage yet. Is it in production with referenceable live customers today, or still early access? If it's live, we raise it to strong and build head-to-head agentic-AI queries against Blue Yonder and Manhattan — a comparison nobody in this category has locked down yet. (3) Are these the right twelve? Two structural checks: is anything missing that buyers ask about — yard management, returns and reverse logistics, or sustainability and freight emissions reporting? And should Core Warehouse Management Execution and Configurability be tested as one capability or two — do buyers experience configurability as a separate thing they're buying, or as the reason the WMS fits?
12 pain points: 9 high, 3 medium, none low — the buyer-language quotes below are literally how the audit's problem-stage queries will be phrased.
→ Three things to check. (1) Nine of twelve are rated high, which means severity can't rank them. Three quarters of this set carries the same severity, so we cannot tell from the data which problem opens your discovery calls. Which three of the nine do buyers actually lead with — the fragmented WMS/TMS/OMS stack, the frozen customized legacy platform, the labor shortage, the 3PL billing lag, or the freight cost blindness? Those three become the problem-stage queries we run first, and getting the order wrong means the audit's most-scrutinised results test the problem your buyers care about fourth. (2) The numbers in the buyer language are ours, not yours. "Thirty people short," "peak is five times normal volume," "eight months and two consultants into a go-live quoted at four," "signed the client in June and still can't bill them in September" — are those the figures your buyers actually say out loud, or should they be different? Queries get phrased in these words verbatim, so a wrong number is a query nobody would type. Note too that peak volatility, 3PL onboarding lag and the delivery-promise pain all carry medium confidence and came from category listings rather than your own reviews. (3) What's missing? Three that plausibly belong in supply chain execution and aren't here: post-acquisition consolidation (an acquirer inheriting three incompatible WMS instances and needing one operating model), tariff and cross-border trade volatility (duty exposure and classification changes that land on the same team), and returns and reverse logistics cost (the cost per return that nobody modelled when the DC was designed). Do any of those come up?
13 findings across 49 pages fetched as raw server HTML on August 13, 2026 — none critical, 3 high, 5 medium, 5 low, two of which need manual verification. The site is fully server-rendered, so nothing here is a rendering problem.
Actionable Now — Engineering There is no emergency in this section, and that's worth stating plainly first: robots.txt is confirmed present, and GPTBot, ClaudeBot and PerplexityBot sit in their own group with Allow: / and no restrictions — there is no AI crawler blocking to chase. Every page also returned complete server-rendered HTML, so client-side rendering is ruled out. Three high-severity items should go to engineering now. (1) Anchor the ten wildcard Disallow patterns in the User-agent: * group. Because they are unanchored they match anywhere in a path, and 11 sitemap URLs are caught by accident — the WMS Buyer's Guide matches /*-s- only because the apostrophe in "buyer's" slugifies to -s-, and two Gartner analyst-report landing pages match /*-cs- and /*-nps-. Googlebot, Google-Extended, ChatGPT-User and Bytespider all fall into that group. Under a day. (2) Emit sitemap lastmod and BlogPosting dateModified from the real last-edited timestamp. Every one of the 936 English URLs currently carries changefreq weekly and a recent lastmod that does not track the content — a February 2024 post claims an April 2026 modification date. One to three days. (3) Fix the llms.txt double-encoding and serve it as text/plain; charset=utf-8. Every non-ASCII character is emitted twice-encoded, so "Körber" renders as "Körber" in the one file written specifically for AI consumers — and Körber is the string that connects a decade of accumulated evidence to Infios. One to three days. Three further findings — the stale blog corpus, the gated buyer's-guide pages and the case study template — are owned by content and run two to four weeks; they should be scoped now rather than started today.
What we found: The catch-all User-agent: * group in https://www.infios.com/robots.txt carries ten unanchored wildcard Disallow patterns intended for campaign and form-confirmation URLs: /*-typ, /*-typ-, /*-ps-, /*-6s-, /*-s-, /*-d-, /*-nps-, /*-cs-, /*-popup and /GTC. Because they are unanchored, they match any URL containing those character sequences anywhere in the path. Cross-checking the 936 URLs in sitemap-en.xml against the patterns, 11 URLs match. The most damaging is https://www.infios.com/en/knowledge-center/supply-chain-resources/wms-buyer-s-guide, which matches /*-s- solely because the apostrophe in "buyer's" is slugified to -s-. Two Gartner analyst-report landing pages (gartner-fap-market-guide-cs-na, gartner-fap-market-guide-nps-na) match /*-cs- and /*-nps-. The named AI crawlers (GPTBot, ClaudeBot, PerplexityBot, CCBot) sit in a separate group with Allow: / and no Disallow lines, so they are unaffected; the pages are blocked for every crawler that falls into the * group, which includes Googlebot, Google-Extended, ChatGPT-User and Bytespider.
Why it matters: Googlebot is the crawler behind Google AI Overviews and AI Mode, and ChatGPT-User is what ChatGPT uses when it browses in response to a live question. A buyer's guide is exactly the asset an AI assistant reaches for when answering "how do I choose a WMS" — and it is the one page in this inventory that is unreachable to those two crawlers. The split-group structure makes this invisible to a casual check: testing GPTBot or ClaudeBot against the URL returns 'allowed', so the breakage only surfaces if you test the * group specifically.
Recommended fix: Anchor the patterns so they match only the URLs they were written for — e.g. Disallow: /*-typ$ for thank-you pages, and move campaign variants under a dedicated prefix (/lp/) that can be disallowed with a single safe rule. At minimum, rename the buyer's guide slug to wms-buyers-guide (no apostrophe) and drop the bare /*-s- and /*-d- patterns, which are too short to be safe on any multi-word slug. Re-test the affected URLs with the robots.txt tester in Google Search Console after the change.
What we found: Every one of the 936 URLs in sitemap-en.xml carries <changefreq>weekly</changefreq> and a recent <lastmod>, but those timestamps do not track the underlying content. Comparing sitemap lastmod against the publish date the page itself renders and declares in its BlogPosting JSON-LD: /blog/logistics-software-build-vs-buy was published 2024-02-27 and claims lastmod 2026-04-09 (a 26-month gap); /blog/5-signs-youve-outgrown-your-current-wms was published 2025-06-17 and claims lastmod 2026-03-20; /blog/the-great-unraveling-erp-vs-order-management-systems was published 2025-05-28 and claims lastmod 2025-09-14. The BlogPosting dateModified matches the sitemap lastmod in each case, so the CMS is stamping both fields on any republish. Separately, 25 of the 47 product and solution pages we fetched share one of two bulk lastmod dates (2026-08-06 and 2026-05-19), which is the signature of a template or CMS-wide republish rather than 25 independent content edits. No page returns a Last-Modified HTTP header.
Why it matters: Crawlers use lastmod to decide what to recrawl and how recent a document is. When the value is systematically inflated it stops being usable, and the pages that genuinely were updated lose the freshness credit they earned — the signal gets discounted for the whole domain, not just the inaccurate URLs. It also means the site's own structured data currently tells an AI assistant that a February 2024 article was modified in April 2026.
Recommended fix: Emit <lastmod> and BlogPosting dateModified from the content's actual last-edited timestamp, and suppress the update when a republish does not change the rendered body. Drop the blanket weekly changefreq — a value that is identical across 936 URLs carries no information. Consider enabling Last-Modified response headers so crawlers have a second, independent signal.
What we found: We pulled the rendered publish date from a random sample of 24 of the 265 English blog posts. 22 of 24 (92%) were published more than 12 months ago, 23 of 24 (96%) more than 6 months ago, exactly one fell inside 90 days, and none inside 30 days. The sample ranges back to 2019 (from-factory-to-footwear, 2019-09-30) and includes posts on cloud computing (2020-03-16) and COVID-era disruption (2020-06-16) still published without any deprecation notice. Within the inventory itself, 9 of the 10 content-marketing pages score at or below 0.2 on freshness. Pages display a single publish date only — no post carries a separately rendered 'last updated' date, so a reader or crawler has no way to see that a 2024 post was reviewed since.
Why it matters: Blog and comparison content is where informational and evaluation queries get answered, and it is the content class where recency weighs most heavily in what AI assistants choose to cite. A corpus where the visible date on nearly every post is over a year old competes badly against vendors publishing and re-dating against the same queries — and the 2019-2020 posts on cloud migration and COVID disruption actively misrepresent Infios as a stale source when they surface. This is compounded by the lastmod finding above: the site's freshest-looking signal (the sitemap) is the one that is not true, and the true signal (the rendered publish date) is the stale one.
Recommended fix: Triage the blog into three buckets: refresh-and-re-date the posts that still map to live buyer questions (WMS selection, 3PL onboarding, freight audit, peak volume); consolidate near-duplicate posts into the pillar pages; and retire or noindex the pre-2022 posts that no longer reflect the product or the market. Render a visible 'Last updated' date distinct from the publish date, and populate BlogPosting dateModified from that same field once it is trustworthy.
What we found: https://www.infios.com/en/supply-chain-solutions/intelligent-supply-chain-execution-old returns HTTP 200 with a body that is byte-for-byte identical to the live page at /intelligent-supply-chain-execution (both 57,167 bytes; identical extracted-text hash). It serves <meta name="robots" content="index, follow"> and is listed in sitemap-en.xml with lastmod 2026-08-07. Its canonical tag does correctly point at the live URL, which contains the damage — but the sitemap is simultaneously telling crawlers to go index a URL the page itself disclaims.
Why it matters: The canonical will resolve this for most crawlers, so this is not an emergency. It is worth fixing because it sends two contradictory instructions about the same URL, and because a -old suffix in a live sitemap suggests the deprecation process leaves retired URLs behind. If that pattern repeats as the product pages are revised, the contradictions accumulate.
Recommended fix: Remove the -old URL from sitemap-en.xml and 301-redirect it to /en/supply-chain-solutions/intelligent-supply-chain-execution. Add a publishing rule that excludes URLs matching a retirement suffix from the sitemap automatically.
What we found: Five commercially positioned pages in the inventory fall below 0.4 on content depth because the substance sits behind a download gate or a link grid rather than on the page. /knowledge-center/supply-chain-resources/wms-buyer-s-guide renders 55 words of body copy and two H2s; /knowledge-center/supply-chain-resources/the-ultimate-tms-buyers-guide renders 43 words. The /supply-chain-challenges hub renders 259 words, /supply-chain-solutions renders 375 words, and /supply-chain-solutions/gamification renders 301 words — the last of which is a full product page. The gated pages follow the pattern across the ~160-page supply-chain-resources section: a title, a one-sentence abstract, and a Download button.
Why it matters: The gated assets are the ones written for exactly the query an AI assistant fields most often in this category — 'how do I choose a WMS', 'what should a TMS RFP cover'. As published, the crawlable page says only that a guide exists. Nothing in it can be cited, so the citation goes to whichever vendor or analyst put the equivalent reasoning on the open page. The hub pages have the same problem in a milder form: they are routing furniture, not answers.
Recommended fix: Publish an ungated summary of each buyer's guide on the landing page — the evaluation criteria, the trade-offs, the questions to ask — and keep the full PDF behind the form. That preserves the lead capture while giving crawlers something to quote. For the hubs and the Gamification page, add the two or three paragraphs of substantive treatment the topic warrants above the link grid.
What we found: The three case studies we fetched carry only Organization and BreadcrumbList JSON-LD — no Article, NewsArticle, or other content type — and each uses exactly two content headings, Challenge and Solution, with no H3 subdivision and no rendered date. The pages do contain specific, quotable outcomes: FabFitFun cut cost per order 66% and saved $10M+ with training dropping from two days to two hours; Do it Best cut routing lead time from 7 days to 1 and moved 40,000 paper invoices a year to electronic; Titan Brands cut backorders 70% and raised customer satisfaction 20%. But at 272-318 words with generic section labels, those numbers sit in passages that carry no topical label an extractor can key on.
Why it matters: Named-customer outcome data is the most defensible evidence Infios has, and it is what an AI assistant needs to justify recommending one vendor over another. Challenge/Solution tells a retrieval system nothing about which challenge or whose solution, so the passage has to be understood from surrounding context rather than standing alone — which is precisely the condition under which a passage does not get cited.
Recommended fix: Rewrite the H2s to name the outcome and the customer ('How FabFitFun cut cost per order 66% across four annual peaks'), add H3s for the specific problem and the specific mechanism, and add Article JSON-LD with datePublished and an articleBody that reflects the actual page text. This is a template change that applies once across all 67 case studies.
What we found: On the blog posts we inspected, the BlogPosting JSON-LD articleBody property contains the page's meta description rather than the article body. On /blog/5-signs-youve-outgrown-your-current-wms, the 849-word article is represented in structured data as the single sentence 'Discover 5 clear signs your WMS is holding you back - and what steps to take to upgrade for scalability, efficiency and integration.' The rest of the BlogPosting object is well-formed — headline, datePublished, author, publisher, mainEntityOfPage, image and inLanguage are all populated correctly.
Why it matters: articleBody is the property that declares what the article actually says. A consumer that trusts the structured data over the rendered HTML — which is the cheap path, and the one several extraction pipelines take — sees a one-sentence teaser where an 849-word argument exists. The visible content is still crawlable, so this is a lost opportunity rather than a blocked page, but it means the site is understating its own best long-form content in the exact field built to advertise it.
Recommended fix: Populate articleBody from the rendered article text, or omit the property entirely — an absent articleBody is better than one that contradicts the page. Verify the fix on a post with the Schema.org validator and Google's Rich Results Test.
What we found: https://www.infios.com/llms.txt exists and is referenced from robots.txt — ahead of most of the market. Two problems with the file as served. First, every non-ASCII character is double-encoded: 'Körber' is emitted as the bytes 4B C3 83 C2 B6 72..., which renders as 'Körber', and em-dashes render as 'â€"'. The file is valid UTF-8, so it is not a transport problem — UTF-8 output has been re-encoded as UTF-8 a second time before writing. It is served as content-type: text/plain with no charset parameter, which removes the one hint a consumer could use to recover. Second, the file is 394 KB across 1,663 lines and appears to enumerate the site rather than curate it, including the full blog archive.
Why it matters: llms.txt is read by exactly the consumers this audit is about, and 'Körber' is the legacy brand name that carries most of Infios's accumulated review and analyst evidence — so the mangled characters land on the single string where entity resolution matters most. The size is a secondary issue: the format's value is being a short, curated map of what matters, and a 394 KB dump of every blog post inverts that.
Recommended fix: Fix the encoding at generation time (write the string once as UTF-8, do not re-encode) and serve as text/plain; charset=utf-8. Then cut the file down to the pages that answer buyer questions — the solution pages, the pillar 'what is' pages, the case studies, and the top evaluation content — rather than the full archive.
What we found: All 49 pages fetched emit rel="alternate" hreflang links for en, de and fr with correctly localised paths (for example the WMS page maps to /de/supply-chain-software-loesungen/warehouse-management-systems-wms and /fr/logiciels-supply-chain/systeme-de-gestion-dentrepot-wms). No page declares hreflang="x-default". Separately, the site is served under three language sitemaps totalling 1,645 URLs (en 936, de 368, fr 341).
Why it matters: Without x-default there is no declared fallback for a visitor or crawler whose locale matches none of the three, so the choice is left to inference. This is a best-practice gap rather than a visibility blocker — the per-language annotations themselves are correct and reciprocal.
Recommended fix: Add <link rel="alternate" hreflang="x-default" href="{english URL}"> to the existing hreflang block in the page template.
What we found: Neither https://www.infios.com/ nor https://www.infios.com/en appears among the 936 <loc> entries in sitemap-en.xml. The root domain redirects correctly (infios.com → 301 → https://www.infios.com/) and the homepage itself returns 200 with a complete server-rendered body, so it is reachable and will be crawled from external links regardless.
Why it matters: Low impact in practice — the homepage is the most-linked URL on any site and does not depend on the sitemap for discovery. It is worth correcting because a sitemap that omits the root suggests the generator is scoped to a CMS subtree rather than to the published site, and the same omission may be dropping other template-level pages we did not detect.
Recommended fix: Add the homepage to sitemap-en.xml (and the /de and /fr roots to their sitemaps), and confirm the generator's page-selection query covers the site root rather than only CMS content nodes.
What we found: Two of the 49 pages emit icon-font ligature names as extractable text. /en/supply-chain-solutions/gamification renders the strings 'procedure', 'productivity', 'assignment_turned_in', 'turnover' and 'exercise' interleaved with its challenge bullets, and /en/supply-chain-solutions/warehouse-management-systems/improve-real-time-inventory-accuracy renders 'warehouse'. Both are the only two pages in the inventory that load a Material Symbols icon font. The strings are invisible to a sighted browser user — the icon glyph replaces them — but they are present in the DOM text and in anything that extracts the page as text.
Why it matters: On the Gamification page the leaked strings sit directly inside the challenge list, so a passage that should read as a coherent claim about turnover and productivity is interrupted by five stray tokens. That page already has the thinnest body copy of any product page in the inventory (301 words), so the pollution is a meaningful share of what an extractor sees. This is narrow — two pages — but cheap to fix.
Recommended fix: Mark the icon spans presentational: <span class="material-symbols-outlined" aria-hidden="true" translate="no">. Better, replace them with inline SVG on these two templates so no text node exists at all.
The following items could not be assessed through our analysis method (raw server HTML, no JavaScript execution). We recommend your engineering team verify these manually before the validation call.
What to check: We parsed JSON-LD directly from the served HTML and recorded the types present on every page: BreadcrumbList and Organization sitewide, Service on 34 solution and product pages, FAQPage on 27, BlogPosting on the 4 blog posts. Spot-checking the FAQPage blocks showed well-formed Question/acceptedAnswer pairs with substantive answer text. What we did not do is validate required-property completeness for each type against Google's own requirements, or confirm the markup passes without warnings — we checked what is present, not whether every emitted object is eligible for the rich result it is aiming at. Worth stating plainly: schema coverage on this site is a genuine strength and scored 0.83 across the inventory. Confirming eligibility protects an investment that has already been made.
Recommended action: Run one representative URL per type — a Service page, an FAQPage, a BlogPosting, the Organization block — through Google's Rich Results Test and the Schema.org validator, and check the Enhancements reports in Search Console for warnings at scale.
What to check: We fetched every page as raw HTML and confirmed the site is server-rendered: all 49 pages returned complete body content, headings, meta tags and JSON-LD in the initial response, with text-to-markup ratios of 0.06-0.30 and 43-2,521 words of main-content text before any script runs. Client-side rendering is therefore not a risk here. We did not run the pages in a browser, so we cannot confirm whether any interactive component — the routing and claims calculators under /transportation-management, the embedded video modules, the tabbed capability panels — contributes content that exists only after hydration. The residual question is narrow: if a tabbed panel or calculator holds substantive copy that only appears on interaction, that copy is invisible to crawlers even on a server-rendered site.
Recommended action: Load three or four representative pages with JavaScript disabled — the WMS page, the Transportation Management page with its calculators, and a case study — and compare against the rendered view. Anything present only with JS on should be moved into the server response.
Scoring Caveat Two limits on the numbers above. Sample size: 49 pages were analyzed against 936 URLs in sitemap-en.xml — roughly 5% of the English tree, weighted toward product, solution and knowledge-center pages. Findings that name a sitewide pattern (the lastmod stamping, the BlogPosting articleBody, the hreflang block) were confirmed on every page in the sample and are template-level, so they generalize; the page-count figures are inventory counts and should be read as such. Freshness coverage: only 10 of 49 pages could be scored. All 38 product and commercial pages and the 1 structural page carry no detectable date in any form — no visible date, no <time> element, and a sitemap lastmod the finding above shows is untrustworthy. Those 39 pages are recorded as unscored rather than as fresh or stale, so the 0.22 weighted figure reflects only the 10 content-marketing pages we could assess. Once real dates are published, expect the picture to change in both directions.
Why Now
The full audit measures how AI assistants answer the questions your buyers actually ask — from problem-stage phrasings like "my WMS, my TMS and my order system all tell me a different story about the same order" and "we customized ourselves into a corner and every upgrade quote looks like a re-implementation", through capability questions like "WMS with multi-client 3PL billing" and "warehouse management that orchestrates AMRs and conveyors", to evaluation-stage phrasings like "Infios alternatives", "Manhattan vs Blue Yonder vs Infor" and "MercuryGate alternatives". You'll see exactly which of those return answers naming Manhattan Associates, Blue Yonder, SAP, Oracle, Infor or Softeon but not Infios — and, given how much of the third-party web still says Körber, HighJump or MercuryGate, which answers are about you without ever using your name. Fixing the Layer 1 findings first means the audit measures a baseline you've already improved rather than one you'll wish you had.
45–60 minutes. We walk through this document together, settle the open questions in the Pre-Call Checklist, and lock the inputs the query set is built from.
We generate buyer queries from the validated personas, competitors, features and pain points, then run them across the selected AI platforms and capture every response.
Visibility analysis, competitive positioning against the primary set, and a three-layer action plan prioritized by which gaps actually cost you citations.
Start Now — Engineering Five Layer 1 items your engineering team can begin before the validation call. (1) Anchor the ten wildcard Disallow patterns in the User-agent: * group of robots.txt — use $-anchored forms such as /*-typ$, drop the bare /*-s- and /*-d- rules, move campaign variants under a single /lp/ prefix, and rename the buyer's-guide slug to wms-buyers-guide; then re-test the 11 affected URLs with the robots.txt tester in Search Console. Under a day. (2) Emit <lastmod> and BlogPosting dateModified from the real last-edited timestamp, suppress the update when a republish doesn't change the rendered body, drop the blanket weekly changefreq, and consider enabling Last-Modified response headers. One to three days. (3) Fix the llms.txt double-encoding — write the string once as UTF-8, serve as text/plain; charset=utf-8, and cut the 394 KB enumeration down to the pages that answer buyer questions. One to three days. (4) Populate BlogPosting articleBody from the rendered article text, or omit the property entirely, and verify on one post with the Schema.org validator. One to three days. (5) Three sub-day cleanups that fit in the same sprint: remove the -old URL from sitemap-en.xml and 301 it to the live page, add the homepage and the /de and /fr roots to their sitemaps, add hreflang="x-default" to the template hreflang block, and mark the Material Symbols spans aria-hidden="true" translate="no" on the two affected pages. These don't depend on the rest of the audit and will improve your baseline visibility before we even measure it. Also worth scheduling this sprint: the two verification items above — the Rich Results Test pass and the JavaScript-disabled render check — both under a day. One thing you do not need to chase: GPTBot, ClaudeBot and PerplexityBot are confirmed allowed with no restrictions, and every page is fully server-rendered. That setup is worth preserving as the site changes.
Two jobs before we meet. The questions on the left require your judgment — no one knows your business better than you. The engineering tasks on the right don't require the call at all.