Technical SEO · DELIVERED
Technical SEO Audit for Pipewright
Pipewright launched on a build platform, and that address still answers. It redirects every page to your real domain — but with a 302, which tells Google the move is temporary, so the ranking credit stays on a domain you do not own. Your homepage's structured data still names that old address as your business, and robots.txt points search engines at a sitemap hosted there too. Separately, and costing more than it looks: two of the four AI answer engines are disallowed in robots.txt, and eleven pages have no text in the raw HTML. Your sitemap, certificate and internal links are clean.
→ to advance
Pages crawled
34
26 listed in the sitemap
Technical health
49/ 100
25 signals measured
Non-indexable pages
3
Blocked from search
AI-unreadable pages
11
No text in raw HTML
Scope · what was measured
What this audit is based on.
Signals measured
- 34 pages crawledEverything we can reach, across the domain and every subdomain. Pages that robots.txt asks search engines to skip are crawled too: an audit has to see a page to tell you it is blocked.
- Raw HTML and rendered pageEach page fetched twice, once without JavaScript and once with it, so every finding can say which of the two it was measured on.
- robots.txt and sitemap.xmlFetched directly from each host, and compared against each other.
- Redirects recorded, never followedThe status code and destination are captured at each hop, so a redirect cannot be mistaken for a second live site.
- https://usepipewright.com/app/ could not be reachedsits behind a login, so there is nothing we can fetch
Pressfit tooling
- Full-domain crawl coverageEvery reachable page across the domain and its subdomains, including paths robots.txt disallows — an audit has to see a page to report that it is blocked.
- Raw + rendered captureEach page captured twice, before and after JavaScript executes, so every finding states which of the two states it was measured against.
- Redirect-chain fidelityStatus codes and destinations recorded hop by hop, never followed. Following a redirect is how one host gets reported as two live sites.
- Crawler + agent parityrobots.txt, the sitemap and each page’s structured data resolved the way a search crawler reads them and again the way an AI agent does, which are not the same resolution.
Scorecard · site health by area
How your site scored.
13
Good
7
Needs work
5
Broken
Readability and speed
3 broken · 4 need work
Host and address
1 broken · 1 need work
Crawlability and indexing
1 broken · 1 need work
Duplication
1 need work
Connection and server
Nothing to fix
Backlinks
Nothing to fix
Sitemap
Nothing to fix
HTTPS and certificate
Nothing to fix
Internal links
Nothing to fix
404 handling
Nothing to fix
Duplicate content
Nothing to fix
Checks · Host and address
Does your site have one address?
Redirect from the old build address
Brokendemo-co-scale.buildhost.app forwards every page to the matching page on usepipewright.com. It says the move is temporary (302), not permanent (301).
Google treats a temporary move as one you might undo, so it keeps the old address on file and leaves the credit there. A permanent one tells it to move the credit across for good.
Measured on the raw HTML
34 pages affected
Hosts serving this site
Needs workTwo: your own, and the old build platform address, which answers and then forwards. It does not serve any pages of its own.
Forwarding is the right behaviour. It is only the temporary wording above that stops the credit moving with it.
Measured on the raw HTML
Canonical tags
PassNine pages have no canonical tag. The seventeen blog posts have one, but as a relative path (/blog/post-name) rather than an absolute URL. Both resolve correctly today.
A relative canonical resolves against the page it sits on, so on your domain it points at your domain. It would only go wrong if the old build address started serving pages instead of redirecting — which it does not. Worth tidying, but it is costing you nothing, and we are not going to tell you to fix something that works.
Measured on both the raw HTML and the rendered page
26 pages affected
www redirect
Healthywww redirects to the main address with a 301 (permanent).
www.yoursite.com and yoursite.com count as two different sites to Google. One has to permanently redirect to the other, or the site competes with itself and each half gets less credit.
Measured on the raw HTML
http to https redirect
Healthyhttp redirects to https with a 301 (permanent).
Anyone arriving on the insecure address has to be moved permanently to the secure one, or the site exists at two addresses again.
Measured on the raw HTML
Redirect loops
HealthyNone.
Two pages redirecting to each other. The browser gives up and shows an error, so nobody can read the page at all.
Measured on the rendered page
If your site loads at more than one web address, Google treats them as separate sites and splits your ranking between them. Fix this first — most other work is worth less until it is settled.
Checks · Crawlability and indexing
Can search engines reach every page?
robots.txt and sitemap
Brokenrobots.txt declares your sitemap as https://demo-co-scale.buildhost.app/sitemap.xml — the old build address, not yours. It also disallows three paths, all correctly.
robots.txt is the first file a search engine reads, and the Sitemap line is how it finds the full list of your pages. Yours points at another host, so your own sitemap — which is complete and correct — is never fetched.
Read from robots.txt
Pages shut out of search results
Pass3 of 34 pages cannot be indexed: two are disallowed in robots.txt and one thank-you page carries a noindex tag. We crawled all three anyway, so we can confirm what is on them. All three look deliberate.
Nothing here is a fault. It is recorded so the number is not mistaken for a problem later.
Measured on the rendered page
3 pages affected
Internal linking and click depth
Needs workTwo orphan pages — nothing on your site links to /platform/integrations or /blog/why-we-rebuilt. The deepest page is 4 clicks from the homepage.
If none of your own pages link to a page, you are telling Google it does not matter. It is the cheapest signal you have and it costs nothing to send.
Measured on the rendered page
2 pages affected
hreflang
Not applicableThe site serves one language in one region, so hreflang does not apply. Considered and ruled out, not skipped.
Measured on the rendered page
Missing pages
HealthyA page that does not exist correctly returns a 404.
If a missing page returns a normal 200 instead of a 404, Google indexes endless addresses that were never real content.
Measured on the raw HTML
Broken internal links
HealthyNone — no links on your site point at a 404.
A link to a page that no longer exists wastes the visit and tells Google the site is not maintained.
Measured on the rendered page
If Google cannot reach a page, or is told to skip it, that page earns you nothing — however good it is. This is about whether your pages can be found at all.
Checks · Connection and server
Is the connection secure?
HTTPS
HealthyEnforced across the whole site.
Without it the browser puts a warning in front of your site, and most people leave when they see one.
Measured on the raw HTML
SSL certificate
HealthyValid until 2027-03-14.
The day it expires, browsers block the site with a full-page warning. It is the fastest way to lose every visitor at once, and it is easy to forget.
Measured on the raw HTML
Directory browsing
HealthySwitched off, so nobody can list the files on your server.
When it is left on, anyone can browse the files in a folder, including ones nobody meant to publish.
Measured on the raw HTML
If a browser cannot verify your site, it warns visitors before they ever see it. And a server set up wrongly will show strangers files you never meant to publish.
Checks · Readability and speed
Can web crawlers and bots read the page?
AI crawler access
BrokenClaudeBot and PerplexityBot are disallowed in robots.txt. GPTBot, CCBot, Google-Extended are allowed.
robots.txt allows anything it does not explicitly disallow, so these are deliberate lines in the file. They are the entire reason those engines cannot cite you, and because it never shows up in your Google rankings it can sit unnoticed for years.
Read from robots.txt
2 pages affected
llms.txt
Needs workNot published.
A proposed file that tells AI crawlers which pages matter and how to read them. It is optional and not yet a standard, so its absence is not a defect — but it is cheap to add and worth having if you want to be cited.
Measured on the raw HTML
Text in the raw HTML
Broken11 of 34 pages have no text in the raw HTML. It appears only after JavaScript runs.
Google runs JavaScript before it reads a page, so this never shows up in your rankings. Most AI crawlers do not — they read the raw HTML and move on, and find nothing.
Measured on both the raw HTML and the rendered page
11 pages affected
Content parsing
Needs work23 of 34 pages parse cleanly. 2 have no main title and 11 have almost no body text once parsed.
A page that yields no title and no body text when parsed gives an answer engine nothing usable, whatever it looks like in a browser.
Measured on the rendered page
11 pages affected
Heading structure
Needs work3 pages have no H1 and 5 have headings out of order. 2 carry FAQ markup, and 14 headings are phrased as a question.
An answer engine lifts a passage, and headings are how it decides where a passage starts and stops. A page with no H1 gives it nothing to anchor to.
Measured on the rendered page
8 pages affected
Structured data
BrokenThe Organization and WebSite blocks on your homepage both give the url as the old build address. Three Product blocks are missing the required offers field. 12 errors, 4 warnings.
Structured data is the summary in a page's code that tells machines what the page is. It feeds Google's knowledge panels and it is the first thing an AI crawler reads. Yours names a domain you are moving away from. The missing offers field means those three pages cannot show a rich result.
0 fatal · 12 errors · 4 warnings
Measured on the rendered page
4 pages affected
Core Web Vitals
Needs workLargest Contentful Paint averages 3.4s on mobile against Google's 2.5s threshold, and 4.2s on the platform pages. Cumulative Layout Shift is fine everywhere.
Largest Contentful Paint is the wait between tapping a link and seeing the main thing on the page. Google uses it as a ranking signal, and most visitors leave before four seconds.
Measured on the rendered page
9 pages affected
HTTP/2
HealthySupported.
Lets the browser collect many files at the same time instead of queuing them one after another. Without it, every page is slower for everybody.
Measured on the raw HTML
Against the pages AI engines actually cite
9 of the 9 pages you are cited for serve their text in the raw HTML.
All nine pages you are cited for serve their text in the raw HTML. None of the eleven JavaScript-only pages has ever been cited.
These signals are measured. We cannot prove one causes the other — those pages differ in other ways too, and no engine publishes how it picks.
Google and AI answer engines both read your pages, but not in the same way — Google runs the page’s code first, most AI crawlers do not. This is what each of them can actually see when they arrive, and how long they wait to see it.
Checks · Duplication
Are any two pages competing?
Duplicate titles and meta descriptions
Needs workFour platform pages share one title tag. Six pages share one meta description.
The title and meta description are what someone reads in the search results before deciding to click. When four pages say the same thing, Google picks one and the rest effectively disappear.
Measured on the rendered page
4 pages affected
Duplicate content
HealthyNo two pages score above the near-duplicate threshold. The closest pair still reads as two different pages.
Measured on the rendered page
When two of your pages go after the same search, Google shows one and ignores the other. It does not ask you which one, and it may not choose the one you would have.
Checks · Backlinks
Who links to you?
312
Referring domains
1847
Total backlinks
4.1%
Scoring as spam
Biggest linking domains
- techcrunch.com214
- g2.com168
- producthunt.com131
- saastr.com96
- capterra.com74
Backlink profile
Pass312 domains link to you, across 1,847 links. The biggest are techcrunch.com, g2.com and producthunt.com. 4.1% of the linking domains score as spam.
Backlinks are still one of the strongest ranking signals Google has. A spam share around 4% is normal and needs no action — what matters is a sudden spike, and there is none.
Measured on the raw HTML
Links from other websites are still one of the strongest signals Google uses to decide whether to trust you. Having very few is the most common reason a technically clean site still does not rank.
Host and address · Finding 01 / 5
- Finding 01CriticalEngineeringLow effort
Your old build address redirects with a 302 instead of a 301
What we observed
demo-co-scale.buildhost.app returns HTTP 302 with a Location header pointing at the matching path on usepipewright.com, for all 34 pages tested. (measured on the raw html.)
Recommendation
Change the redirect from 302 to 301 in your build platform's domain settings. The path-for-path mapping is already correct — only the status code is wrong.
Check this yourself
- curl -sI https://demo-co-scale.buildhost.app/pricing
- Expect: A 302 today, with a location line pointing at https://usepipewright.com/pricing. Once fixed, the same command answers 301.
Run it exactly as written. Adding -L follows the redirect and fetches the destination instead, which hides the problem.
Pages affected
- https://demo-co-scale.buildhost.app/
Why it matters
A 301 is permanent and tells Google to move the ranking credit to the new address. A 302 is temporary and tells it to keep the old address indexed and leave the credit there. Every link anyone has built to your launch address is currently working for a domain you do not own.
Estimate · That old address is credited with 41 search terms worth roughly 6,800 searches a month. That is what is sitting on the wrong domain.
Crawlability and indexing · Finding 02 / 5
- Finding 02CriticalEngineeringLow effort
Two of the four AI answer engines are disallowed in robots.txt
What we observed
robots.txt contains User-agent: ClaudeBot / Disallow: / and the same for PerplexityBot. GPTBot and Google-Extended are allowed. CCBot is not mentioned either way. (read from robots.txt.)
Recommendation
Remove both Disallow blocks, unless the block was deliberate. If it was, this closes — but it should be a decision you made rather than a default that arrived with the theme.
Check this yourself
- curl -s https://usepipewright.com/robots.txt
- Expect: No line telling ClaudeBot or PerplexityBot to skip everything.
Run it exactly as written. Adding -L follows the redirect and fetches the destination instead, which hides the problem.
Pages affected
- https://usepipewright.com/robots.txt
Why it matters
Those two lines are the entire reason those engines cannot quote you. It costs nothing to be readable, and because it never shows up in your Google rankings it can sit there for years without anyone noticing.
Crawlability and indexing · Finding 03 / 5
- Finding 03HighEngineeringLow effort
robots.txt points search engines at a sitemap on the wrong domain
What we observed
The Sitemap line in https://usepipewright.com/robots.txt reads https://demo-co-scale.buildhost.app/sitemap.xml — the old build address rather than yours. (read from robots.txt.)
Recommendation
Change that one line to https://usepipewright.com/sitemap.xml. Nothing else in the file needs touching — the three disallowed paths are all correct.
Check this yourself
- curl -s https://usepipewright.com/robots.txt | grep -i sitemap
- Expect: One line reading Sitemap: https://usepipewright.com/sitemap.xml
Run it exactly as written. Adding -L follows the redirect and fetches the destination instead, which hides the problem.
Pages affected
- https://usepipewright.com/robots.txt
Why it matters
robots.txt is the first file a search engine reads on any site, and the Sitemap line is how it finds the full list of what you publish. Yours sends it to another host. Your own sitemap is complete and correct, and is simply never being fetched.
Readability and speed · Finding 04 / 5
- Finding 04HighEngineeringLow effort
Your structured data names the wrong website as your business
What we observed
The Organization and WebSite blocks on your homepage both give the url as demo-co-scale.buildhost.app. 12 errors across four pages, including three Product blocks missing the required offers field. (measured on the rendered page.)
Recommendation
Set both url values to https://usepipewright.com, and add the offers field to the three Product blocks on the platform pages.
Check this yourself
- curl -s https://usepipewright.com/ | grep -o '"url":"[^"]*"' | head
- Expect: usepipewright.com. Today it answers with the buildhost.app address.
Run it exactly as written. Adding -L follows the redirect and fetches the destination instead, which hides the problem.
Pages affected
- https://usepipewright.com/
- https://usepipewright.com/platform
Why it matters
Structured data is the summary in a page's code that tells machines what the page is. It fills Google's knowledge panels and it is the first thing an AI crawler reads. Yours currently names the domain you are moving away from. The missing offers field means those three pages cannot show a rich result at all.
Readability and speed · Finding 05 / 5
- Finding 05HighEngineeringHigh effort
Eleven pages have no text in the raw HTML
What we observed
Fetched without JavaScript, /platform, /pricing and nine others return a shell with no body copy. The same pages render fully once JavaScript runs. (measured on both the raw html and the rendered page.)
Recommendation
Render those eleven pages on the server so the text arrives in the HTML. Every one is a marketing page whose content is the same for every visitor, so nothing is lost by doing it.
Check this yourself
- curl -s https://usepipewright.com/pricing | wc -c
- Expect: Around 4,000 characters today — an empty shell. A page sent out complete comes back around ten times that, with the pricing words in it.
Run it exactly as written. Adding -L follows the redirect and fetches the destination instead, which hides the problem.
Pages affected
- https://usepipewright.com/platform
- https://usepipewright.com/pricing
- https://usepipewright.com/platform/workflows
Why it matters
Google runs JavaScript before reading a page, so this never shows up in your rankings. Most AI crawlers do not — they read the raw HTML and move on, and yours is blank. These eleven include your pricing page and every platform page: exactly the ones you would want cited in an answer.
Estimate · Those eleven pages cover about 3,900 searches a month between them, and not one has ever been quoted by an AI engine.
Prioritization · impact × effort
What to do first.
- 1302 to 301 on the old build address
- 2Unblock ClaudeBot and PerplexityBot
- 3Fix the Sitemap line in robots.txt
- 4Fix the url in your structured data
- 5Server-render the eleven blank pages
Across the bottom is how much work each one is. Up the side is what it is worth. The top-left corner is where to start.
Roadmap · 90 days
The order these want to happen in.
Phase 1 - Days 0-30
Settle which address you are
One address, forwarding permanently, named correctly everywhere — so everything measured after this is measured against the right site
- Change the old build address redirect from 302 to 301
- Point the Sitemap line in robots.txt at usepipewright.com
- Set Organization.url and WebSite.url to usepipewright.com
- Remove the ClaudeBot and PerplexityBot Disallow blocks
Phase 2 - Days 30-60
Tidy what search engines see
Every page distinct, reachable, and complete in its summary
- Give the four platform pages distinct titles and meta descriptions
- Link to the two orphan pages from the platform and blog indexes
- Add the missing offers field to the three Product blocks
Phase 3 - Days 60-90
Readable by anything, quickly
The words arrive with the page, and they arrive fast
- Server-render the eleven JavaScript-only pages, pricing first
- Bring the platform pages under Google's 2.5s Largest Contentful Paint threshold
- Add an H1 to the three pages missing one and fix the five broken heading orders
- Re-measure AI citations 30 days after the pages go server-side
Settle which address you are first. Until that is done, everything else is being measured against the wrong site.
What's next
Want this for your site?
Everything here is sample data for a fictional company. Yours is built from your site, your market and your competitors, and your marketer goes through it with you on a call.