Full technical version · August 2026
Bloomery Home: 106,000 followers and a site Google cannot see
This is the uncut document — the same one a store owner hands to their developer. If you are an owner and just want to understand what is wrong, read the short version in plain language: same content, two minutes.
In brief
A store selling homeware, on a hosted ecommerce platform. A strong Instagram brand — 106,000 followers, over 1,600 posts. Catalog per the sitemap: 513 products and 40 category and information pages.
And yet in organic search the store barely exists: 44 queries with any visibility, roughly 300 impressions and 9 clicks a day, and the only query that brings clicks is the brand's own name.
The headline conclusion: this is not an ad-budget problem. The site is technically unprepared to be found — it has no headings, no descriptions and no product markup, which is to say none of the material search builds a result from. The audience exists, the demand exists, the search infrastructure does not.
What I saw and what I did not. This audit was done from the outside plus access to Search Console. Without GA4, the store admin and order data, some conclusions (real conversion rate, cart behaviour, margin) cannot be verified — and I say so explicitly below rather than adding plausible guesses.
Starting data
| Metric | Value |
|---|---|
| Platform | Hosted ecommerce platform, default theme |
| Site language | One language, one version (no hreflang needed) |
| URLs in the sitemap | 513 products + 40 pages = 553 |
| Queries with visibility in Google | 44 |
| Clicks / impressions (per day) | ≈9 / ≈300 |
| Non-brand clicks | 0 |
| 106,000 followers, 1,617 posts | |
| TTFB | 0.22–0.57 s — fine |
1. Technical health
1.1. A JS check before every first visit
The first request to any page returns not the site but a 518-byte page carrying a bot-protection script. The script spins an empty loop for 800–1300 milliseconds, sets a cookie and reloads the page.
For someone arriving from Instagram that is a second to a second and a half of blank white screen plus a full reload — at exactly the moment the "stay or close" decision has not been made yet. For search and AI crawlers it is a risk: some of them do not execute JS and will see only those 518 bytes.
What to do: establish why this protection is on, and at minimum add Googlebot, Bingbot and the AI crawlers to an exception list. Better still, switch it off entirely if there is no real scraping problem.
1.2. Images all load at once
The kitchen category page carries 64 catalog images and not one has loading="lazy". The browser fetches all 64 immediately, including the ones below the fold.
The files themselves are done well — around 6 KB on average, and the platform generates thumbnails correctly. So this is a request-count problem, not a weight problem. WebP is not in use either: the CDN serves JPEG even when the browser explicitly asks for WebP. That would save another 25–30%, but it is a detail next to point 1.1.
1.3. What is fine technically
The sitemap exists, is correctly split into products and pages, and updates daily. robots.txt tidily blocks filters, sorting, search and the cart — competently done. Canonical addresses are in place. Server response time is fine. Breadcrumbs are marked up.
Separately: robots.txt carries Content-signal: ai-train=no. That is a deliberate refusal to allow model training. The decision is yours, but it is worth knowing you made it — often this is just a CDN default nobody read.
2. SEO and visibility
2.1. There is not a single meta description
Not on the homepage, not on categories, not on product pages. Google then cuts its own snippet out of the page — usually a poor one. This directly affects click-through: 300 impressions a day at zero clicks is partly about exactly this.
2.2. The homepage has no headings at all
No <h1>, no <h2> — none. As far as search is concerned the homepage has no subject: there is no text explaining what this store is and what it sells. Categories and product pages do have an <h1>, but <h2> appears nowhere — so there is no content structure.
2.3. There is no Schema.org product markup
The product page has no structured data whatsoever — no Product, no Offer, no price, no availability. The homepage has only WebSite and SearchAction.
The consequences are specific: no price, availability or stars in the search results; Merchant Center cannot reconcile its data against the page; ChatGPT and Google AI Overviews have nothing to assemble an answer about your product from. For a store this is the most expensive of the technical gaps.
2.4. Titles eaten by the brand
Every title ends with "— Bloomery Home — online store of cosy homeware". That is 58 characters on every page. Google shows around 60 — so almost nothing is left for the product name itself, and every page looks identical in the results.
What to do: shorten the suffix to "— Bloomery Home" and give the freed space to the attributes people actually search on: size, colour, material.
2.5. Category addresses do not match their names
A category called "Kitchen" lives at /tableware/. Mismatches like that are a signal the catalog structure was not built from demand. Judging by the Search Console queries, demand exists for specific objects: "salt and pepper set", "lemon dish", "paper towel holder", "doormat".
Those are not general-purpose categories — they are ready-made subcategories, each deserving its own page and its own text. Right now no such pages exist, and the queries land on general categories with nothing to compete on.
2.6. What Search Console says
All 44 queries have microscopic impression counts (2–6) and zero clicks, except the brand one. That is the classic picture of a site Google has indexed but does not consider relevant to anything specific. Together with 2.1–2.3 it is consistent: no descriptions, no headings, no markup — and therefore no grounds to rank it.
The good news: the queries that do get impressions are commercial and precise. The demand exists; there are simply no pages for it.
2.7. AI-search visibility
llms.txt returns a 404. Together with the missing Product markup and ai-train=no, that means the store is effectively invisible in AI search. For a gifts and decor niche, where a lot of searching is descriptive ("what to give for a housewarming"), that is a lost channel.
3. Conversion
Here I am limited: without GA4 and order statistics, real conversion cannot be assessed. What is visible from outside:
- The product page has an
<h1>and breadcrumbs — the basics are right. - There is no product description that could be called text — which matters both for search and for the decision to buy.
- The delay from point 1.1 hits Instagram traffic specifically, which is your main channel.
To go further I would need: access to GA4, the "product page → cart → order" funnel, and a dozen session recordings. Without those, any conversion advice is guesswork.
4. Analytics
Google Analytics and the Meta Pixel are both present on the site; the code shows the platform's ecommerce events configured (cart creation, add to cart). That is good — the basic infrastructure exists.
What has to be checked from inside and is not visible from outside: whether this is GA4 or still Universal; whether purchase events arrive with a real amount attached; whether Merchant Center is connected and whether products are being rejected there over a price mismatch (see 2.3).
5. Operations
From outside, payment and shipping look to be within the platform's standard capabilities, and the "Payment and delivery" and "Exchange and returns" pages are in place. Order handling, receipts and CRM cannot be assessed without admin access.
One question I would put to the owner first: how many orders a month arrive through Instagram DMs, and who processes them by hand? At 106,000 followers that is usually the main hidden cost line.
Priorities
Sorted by impact against effort. The top three items will deliver more than everything else combined.
| Problem | Impact | Effort | What to do |
|---|---|---|---|
| No meta descriptions | high | low | Description templates for categories and product pages. One evening of work. |
| No Product markup | high | low | Enable or hand-write Schema.org in the product template: price, availability, brand. |
| JS check on the first visit | high | low | An exception for search and AI crawlers, or switch it off entirely. |
| The brand eats the title | medium | low | Shorten the suffix; a title template carrying product attributes. |
| No subcategories for real demand | high | medium | Do the keyword research, create 15–25 subcategories with text. |
| No headings on the homepage | medium | low | An H1 and an H2 structure describing the range. |
| No product descriptions | high | medium | 513 descriptions from the attributes — realistically a week's work via LLM. |
| 64 images with no lazy-load | medium | low | Enable deferred loading in the theme. |
| No llms.txt, ai-train=no | medium | low | Add llms.txt; make the AI policy a deliberate decision. |
| No WebP | low | low | Enable conversion at the CDN. Minus 25–30% weight. |
| Category URLs ≠ names | low | medium | Bring them into line with redirects. Not urgent. |
What I think about this store overall. The most expensive part has not been spent: the brand and the audience are already built, and that is the part no budget can buy. The gaps are technical, and almost everything in the top half of the table closes within a few days of work. This is the rare situation where the effort-to-result ratio genuinely favours the owner.
What happens next
The same document for your store: 48 hours, free. If you then do everything yourself and order nothing from me, that is a normal outcome — half the items above really can be done in-house.