flexecharge.com, read by a crawler before the rebuild

The short version

Your sitemap describes a site that no longer exists; the real site discovers itself only through links. Four pages carry a canonical tag that points at a URL redirecting back to them, and two English case studies declare the French copy as canonical. Half of the French mirror the sitemap promises returns 404, and most of the half that exists tells Google it is a duplicate of the English page. The homepage declares two different companies called FLEXECHARGE. Those four are the before-build items. The rest is build hygiene and content metadata, listed in order below.

What is already right, so nobody re-does it: the apex host redirects to www cleanly on every probe, the 404 page returns a real 404, trailing slashes redirect, robots allows the AI crawlers by name, llms.txt exists, every page carries Organization and WebSite JSON-LD and an hreflang set, and the glossary is typed as a DefinedTermSet.

Findings, in the order to act on them

1. 1
   Fix before the build starts

   The sitemap is out of date

   Of the URLs in sitemap.xml, return 200 at the address listed. redirect ( of them through two hops) and return 404. Meanwhile the pages the site itself links to include live pages the sitemap never mentions at any address: blog posts, the podcast episode pages, and the audience pages already published for depot, destination, high-power hubs and energy-provider CPOs. A further live pages are listed only under a retired address that redirects. Two of the dead sitemap URLs are your local and cloud load management technology pages: they return 404 with no redirect while still listed in your sitemap.

   For GoogleA sitemap full of 404s and redirects lowers trust in the file; the pages that matter most this year are discoverable only by crawling links, which is slower and misses orphans.

   For AI answersAnswer engines and their crawlers (OAI-SearchBot, PerplexityBot, ClaudeBot) read sitemaps to find fresh pages. Most of your newer posts and the new audience pages are invisible to that path.

   The 12 dead URLs

   URL	Linked from live pages

   The 27 redirecting URLs

   Listed	Resolves to	Hops

   Do: switch to Webflow's auto-generated sitemap (or regenerate the custom one from the live CMS), remove the 404s, list final URLs only, resubmit in Search Console. Before any redirect plan is written, this file is the inventory it has to be written against.
2. 2
   Fix before the build starts

   Canonical tags that contradict the redirects

   Four live pages carry a canonical pointing at their old URL, and the old URL 301-redirects back to the page. Google is told two opposite things about which address is real. Two English customer cases declare the /fr/ copy as canonical. /products/connect carries two canonical tags. of the live sitemap pages and of the newer pages carry no canonical at all, including three of the four new audience pages. On every probe, the apex host and plain http redirected to https://www in at most two hops, all canonicals found point at www, and none of the live pages links to the apex host. So the apex-versus-www issue you mention is not in the HTML; see the host section below for where to look.

   For GoogleA canonical that points at a redirect is ignored or, worse, consolidates signals onto a URL that no longer serves; the fr canonicals hand English equity to pages that are mostly English anyway.

   For AI answersCitation engines resolve the canonical to decide which URL to show. A loop means the cited address may be the retired one.

   The six canonical conflicts

   Page	Canonical says	What that URL does

   Do: set every canonical to the page's own final URL (Webflow: clear the per-page override, let the global canonical rule take it), one tag per page, and add canonicals to the audience pages before they earn links.
3. 3
   Fix before the build starts

   The French mirror cancels itself out

   The sitemap declares a French alternate for all URLs. of them return 404: every English URL that was redirected or removed lost its French twin without a redirect. Of the French pages that exist, declare the English page as their canonical, declare lang="en", and carry an English title. The newer pages all declare a French alternate too; of eight probed, six exist and two return 404.

   For Googlehreflang says "this is the French version"; the canonical says "this is a copy of the English one". Google follows the canonical, so the French pages are dropped and the hreflang is wasted work.

   For AI answersFrench-language questions about load management for CPOs will be answered from someone else's French pages.

   Do: decide whether French is a maintained locale. If yes: self-canonical on every /fr/ page, lang="fr", French titles, and the same redirect map applied under /fr/. If no: drop the hreflang and the alternates from the sitemap rather than advertising 404s.
4. 4
   Fix in the new homepage

   The homepage declares two different FLEXECHARGEs

   The homepage carries Organization blocks. They disagree on the logo file, the description, the office addresses (Nordre Fasanvej against Rahbeks Allé in Frederiksberg; Erwitter Strasse against Beethovenstrasse in Lippstadt), the contact fields, and the sameAs list (four profiles against two). The heading outline has two h1 elements, because the mobile hero is a second copy of the section with a different headline; h2 elements whose whole text is a digit; an h5 before the first h3; and "But we thrive on challenges" as both an h2 and an h4.

   For GoogleTwo conflicting entities on the home URL weakens the knowledge-panel signal; the heading outline reads as a slide deck, not a document.

   For AI answersEntity clarity is the thing you listed under the solution-page template; it is broken on the one page every engine reads first. A model asked "where is FLEXECHARGE based" gets two answers from one page.

   The two Organization blocks, side by side

   Do: one Organization node, with @id, one logo, both offices, the four sameAs profiles, and every other page referencing it. One h1 rendered once (hide the variant with CSS inside one element, not a second element). Numbers in the challenges list become list markers, not headings.
5. 5
   Fix in the build

   Every article says it was written on the same day, by a person named FLEXECHARGE

   Across the live article pages found, carry datePublished: "Sep 16, 2025", including monthly round-ups about May, June and July 2024. name the author as a Person called "FLEXECHARGE". Dates are written as display strings, not ISO dates. Evergreen explainers are typed NewsArticle; only the five whitepapers are typed Article with an Organization author and an ISO date, which is the pattern the rest should follow.

   For GoogleUnparseable dates lose the date snippet and the freshness signal; a Person who is a company is an E-E-A-T miss.

   For AI answersRecency and authorship are two of the few signals answer engines weight when choosing between near-identical sources. A 2024 round-up stamped September 2025 looks stale-and-mislabelled at once.

   Do: in the Webflow CMS template, bind datePublished and dateModified to the item's date fields in ISO format, type posts as BlogPosting (or Article), author as the Organization or a real named Person with a profile page.
6. 6
   Fix in the templates

   Heading outlines skip levels and count with headings

   of the live sitemap pages skip at least one heading level. headings on pages contain only a digit. of the newer blog posts render two h1 elements. /products/optimise renders its h1 twice.

   For GoogleMinor for ranking, major for the "answer-ready sections" you want: passage ranking and featured snippets lift sections whose heading states the question.

   For AI answersModels chunk pages by heading. A section headed "2" is a chunk with no topic.

   Do: one h1 per page from the CMS title; h2 per answerable question; styled numerals as span or list markers. Put this in the solution-page template once and every page inherits it.
7. 7
   Fix in the templates

   Titles and descriptions with template residue

   One live post's description reads FR - FLEXECHARGE Canonical definition 100% D80 Explore the future…, a template note that shipped. Two posts have the title | FLEXeCHARGE and nothing else. titles end in the brand twice (… | FLEXECHARGE | FLEXeCHARGE). live sitemap pages and newer pages have no meta description, mostly brochures. No page sets og:url.

   Do: one title pattern in the CMS template with the brand appended once; a required description field on brochures and posts; og:url bound to the canonical.
8. 8
   Add in the solution-page template

   Schema stops at Organization and WebSite

   Every page carries Organization and WebSite. Beyond that: page with FAQPage (an event page), page with SoftwareApplication (the HARMON-E platform page), no Product or Service on the product pages, and JSON-LD block that fails to parse (the FLEXBOX page). The glossary's DefinedTermSet is the one page-specific type done right, and it is the kind of page answer engines quote.

   Do: the solution-page template ships with Service or SoftwareApplication (name, provider as the single Organization @id, audience, areaServed), FAQPage bound to the on-page FAQ, and BreadcrumbList. Validate the FLEXBOX block; it is probably an unescaped quote.
9. 9
   After launch, content pass

   Images without alt text

   of image elements on the live sitemap pages have no alt attribute or an empty one. Decorative images should be empty on purpose; customer logos and diagrams should not.

The apex-versus-www question

Every probe of the bare host resolved to https://www.flexecharge.com, so the live HTML is not where the inconsistency lives. The candidates, in the order I would check them: the Search Console property (an apex property and a www property report different pages), external and AI citations that still use flexecharge.com/… and pay two redirect hops from plain http, the new build's publish settings in Webflow (default domain must stay www), and any hard-coded apex URLs in the redesign's mock-ups. Send me what you are seeing and I will say which one it is.

Product and technology pages a consolidation would retire

Your product-based structure today, as the crawl found it: what each URL does now and what it currently answers. The target column needs two things only you have, the legacy list you are consolidating and the pages your AI-citation source says are cited; with those, each row gets a single named destination, no chains.

Rule for the map: one hop, same language, to the page that answers the same question. A technology page should land on the explainer or whitepaper that covers it, not on a product overview; an already-dead page gets a redirect too, because the citations that pointed at it are still out there.

Two things this review covers only as guidance

The interactive European map

There is no map on the live homepage today, so this is a build rule rather than a finding. Render the country list and each country's facts as HTML first (a list or a table, one anchor per country), then let the map be a layer that reads from that HTML. Nothing that should be indexed lives only inside canvas, SVG labels or a JavaScript data file; the map's tooltip text is the same text as the list; keyboard focus moves through countries in reading order.

The solution-page template

• One h1 that names the audience's problem in their words; the first paragraph answers it in two sentences a model can lift whole.
• h2 sections that are questions a buyer asks, each followed by a direct answer before any detail.
• An FAQ block on the page, with FAQPage schema bound to it, not a second copy.
• Service or SoftwareApplication schema referencing the single Organization @id.
• Links out to the specific case study, glossary term and whitepaper that back each claim; links in from those pages.
• Self-canonical, one title, a description written for the page, og:url, and a French alternate only if the French page exists.

What this page cannot see

I did not have the redesign plans, the list of legacy URLs you are consolidating, your Search Console data, or whichever tool tells you which pages AI engines cite. The findings above are what the live site shows a crawler; the plan review proper starts when those four arrive.

Method and self-check

Raw HTML fetched sequentially with a desktop browser user agent and a 0.6 second pause between requests; no markdown conversion, because that strips the script tags that hold JSON-LD. Redirects were followed by hand so each hop is recorded. Per page: status, final URL, hops, canonical tags, meta description, title, heading outline, JSON-LD blocks parsed, hreflang set, word count, internal links, images. The table below recomputes the counts quoted above from the embedded records and compares them with the offline derivation; a mismatch would show here as a fail.