Search Overflow
Scan my site

Receipts

Our own site, checked the way we check yours.

On 24 September 2026 our engine checked our homepage the way it checks yours: 3 issues worth fixing, 2 notes and 2 things it could not check. One of the 3 was our engine’s mistake, not our page’s, and we say so below.

Our homepage, checked

One page, read the way an AI crawler reads it.

Worth fixing
3
Notes
2
Could not check
2

Our homepage, issue by issue

Each issue shows the page and what we found on it, the change, and the re-check. Where the check counted something, the count is shown exactly as it was recorded, and anything not done yet says ‘not run yet’. The third issue worth fixing is further down, because it was our mistake, not our page’s.

worth fixing

The page shows no date an AI assistant can read.

The check said: “The page carries no date a model can read”

Evidence
timestamps_found: 0
The change
A dateModified in our structured data and a visible last-updated line.
@@ JSON-LD @@
+ "@type": "WebPage",
+ "dateModified": "2026-09-24",
@@ visible text @@
+ Page last updated 24 September 2026
Re-check
not run yet

worth fixing

Our structured data links to none of our profiles on other sites, so an assistant cannot tell they belong to the same business.

The check said: “No sameAs links to authoritative profiles”

Evidence
same_as_count: 0authoritative: []
The change
Links from our structured data to our profiles on other sites, once we list them.not made yet
Re-check
not run yet

note

Our structured data carries no review ratings.

The check said: “No Review or AggregateRating schema”

Evidence
The check recorded no count for this one, only what is missing.
The change
None planned. We publish no reviews today, and we will add ratings only from genuine ones.
Re-check
Not needed. Nothing on the page changes.

note

Our homepage carried FAQ markup, which Google no longer uses for rich results.

The check said: “FAQPage markup is deprecated and does nothing”

Evidence
blocks: 1retired_on: "2026-05-07"
The change
We removed the FAQ markup. The questions and answers stay on the page for readers.
@@ JSON-LD @@
- "@type": "FAQPage",
- "mainEntity": [ 6 questions and answers ]
@@ visible text @@
  Questions and answers stay, for readers.
Re-check
not run yet

What it could not check

  • not checked yetRankingsNo Search Console connection, so no ranking data for this site.
  • not checked yetBusiness profilesNo Google Business, Facebook or Instagram profile is connected.

One warning was our engine’s mistake

Our engine warned that the prices in our structured data were empty. They were not. Each plan lists its price directly on its offer, and our product entry points to those offers by reference.

The engine looked for a price only one level down, inside a nested offers field. It never read a price set on the offer itself, and it did not follow the references. So it read every entry as empty.

Any client whose markup is shaped like ours would have got the same wrong warning.

So the fix belongs in the engine, not on our page. Until it ships, the receipt says not made yet.

worth fixing

Our engine read the prices in our structured data as empty. They were not.

The check said: “The schema is there, but the values that matter are missing”

Evidence
empty_blocks: 4product_blocks: 4
The change
Nothing on our page. The fix belongs in our engine.not made yet
Re-check
not run yet

How often assistants name us: not measured yet

We have not yet asked AI assistants how often they name Search Overflow, so this page shows no rate. When we do, every rate will show how many answers it rests on and a 95% confidence range, and none will come from fewer than 7 answers.

AI mentions of Search Overflownot run yet

What we took back

Claims our site made before this page existed that our code did not back. We removed them.

  1. We said“Three labels. No false passes.”Why it wentOur free scan could report a bot as allowed when it had not been able to read the site’s robots.txt file at all.
  2. We said“A change reaches your site only when someone on your side reviews the pull request and merges it.”Why it wentOur code does not guarantee that on every route a change can take. What it does guarantee stays on the site: we never merge.
  3. We said“Supported fixes arrive as a pull request.”Why it wentOpening pull requests automatically is switched off in production today. Our team opens them, or hands you the exact change.
  4. We said“Six AI assistants tracked”, on the plans.Why it wentTracking AI assistants is work our team does by hand. The product does not do it on its own yet.
  5. We said“Do you use AI to generate the scan results? No.”Why it wentIn some of our SEO checks a language model reads your page copy, and when we track AI assistants a model reads each saved answer. Our FAQ now says where.
Our internal issue logEvery issue our own checks raise about this site and its code, open ones included, and the limits of those checks.Show the log

Everything else we check on our own site

Beyond the homepage, we run the same kind of checks on our whole site and our code. Every issue they raise goes into one log. An issue counts as fixed and re-checked only when a second check, not the fix itself, confirms it.

Security and privacy issues are left out of this page on purpose, so it cannot help anyone attack us. We track them in the same log.

Last full pass of every check
not run yet
Last partial pass

Generated 24 September 2026 from our issue log at commit 05645a20eb27. It includes changes not yet committed.

Issues by area

  • Usable without sight, mouse or perfect visionOpen1Fixed0Re-checked0Closed0
  • Answer engine readinessOpen4Fixed0Re-checked0Closed85
  • Backend code healthOpen85Fixed0Re-checked0Closed0
  • Claims against codeOpen10Fixed1Re-checked0Closed2
  • Plain, specific copyOpen12Fixed0Re-checked0Closed0
  • Frontend code healthOpen53Fixed0Re-checked0Closed5
  • Deploy and production operationsOpen16Fixed1Re-checked5Closed0
  • Auditable agency positioningOpen1Fixed0Re-checked0Closed0
  • Our own technical SEOOpen19Fixed0Re-checked0Closed19
  • Regression gate and test safetyOpen1Fixed1Re-checked0Closed0
  • Honest proofOpen1Fixed0Re-checked0Closed0

Fixed means changed and waiting for its re-check. Closed means accepted with a reason, or closed on a second look as a duplicate or not a real problem. Areas with no issues logged are not listed, and that does not mean they are clean.

Still open, by severity

Critical
0
High
13
Medium
51
Low
141
Note
1

Fixed and re-checked

None listed by name yet.

5 of 5 are not listed by name, because their titles name internal files or systems.

When we call our own site done

We stop working through our own log only when all 5 of these hold. Today 0 of 5 hold.

  • not yetTwo full passes in a row, on the same code, find nothing new.
  • not yetNo critical or high-severity issue is open.
  • not yetNo medium or low issue is open.
  • not yetThe full test suite passed on the last full pass.
  • not yetTwo review rounds in a row, each by AI reviewers working separately, score every area at or above its bar.

What our checks cannot see

Every check has limits, and we keep notes on them for each area. These are the first lines of those notes. Notes on security and privacy are left out for the same reason as above.

The checks we run on ourselves

  • A map of the code: every route, check list and version, compared with what the site says.
  • A real browser at desktop and mobile width, loading each page and checking its accessibility and speed.
  • The words on every page, checked against the code for counts, overclaims and repetition.
  • Security checks on our own stack. Their results are not published here.
  • Code health: file size, lint, types, and code nothing uses.
  • Our own SEO and AEO checks, run on our own site: the same ones we run on yours.
  • The full test suite, run offline.

Usable without sight, mouse or perfect vision

  • Only the default dark theme is swept.
  • axe "incomplete" results are not stored.
  • WCAG 2.2 is not run.
  • Interactive states are never exercised: open menus, focus order, focus visibility, keyboard traps, form errors.

Answer engine readiness

  • Allowed is not crawled, and crawled is not cited.
  • The drivers evidence was measured "in a two-document retrieval testbed, not against live answer engines".
  • The seven questions are a judgement, not measured prompts.

Backend code health

  • The architecture tests read imports only.
  • Whether a module used only by tests is planned work or a leftover is an owner decision.
  • Size and lint say nothing about correctness.

Claims against code

  • Production database state cannot be read.
  • Business facts have no code source.
  • Paywalled or bot-blocked sources cannot always be fetched to confirm a figure.

Path to a recorded next step

  • We never submit forms with real emails, so we cannot see whether waitlist, report and contact emails are delivered or how fast a person replies.
  • There is no analytics by design, so conversion rates and drop-off are unknown.
  • The production launch mode is inferred from live redirects.
  • Whether a manual follow-up happens within the time the page promises cannot be observed from here.

Plain, specific copy

  • Tone and voice are subjective.
  • Only English is read.
  • Blog depth and topical authority cannot be measured here.

Five-second test

  • No real buyer sees the page.
  • Screenshots come from headless Chromium at 1440x900 and 390x844 only.
  • Who actually visits is unknown, because the site runs no analytics by design.
  • Whether a buyer finds our "real artefact" convincing is a judgement, not a measurement.

Frontend code health

  • Line count is not responsibility.

Assistant visibility readiness

  • The outcome itself is unmeasured.
  • A model's built-in knowledge of a 2026 brand may lag its training cut-off.
  • Consumer apps personalise answers, and India cannot be sampled from an Indian exit.
  • Mention detection is a model extraction over aliases.

Deploy and production operations

  • Health endpoints report configuration flags, not end-to-end success.

Fast and stable where our buyers are

  • One cold lab load per page.
  • No throttling.
  • Unknown vantage point.
  • No field data.

Auditable agency positioning

  • Whether the category sentence lands with real buyers is untested.
  • Competitor facts are captures from 2026-09-23 and go stale.
  • Competitor products behind login were never seen.

Our own technical SEO

  • No Search Console data for our own property.
  • Crawler behaviour is not observed.
  • The route cap is 60.
  • Rich-result eligibility is Google's decision.

Regression gate and test safety

  • The gate runs each suite once, in default order.
  • next build runs on this machine, not in the Docker image, with a different node version.
  • Pass counts say nothing about assertion strength.

Honest proof

  • Whether named founders and a self-audit persuade real buyers is untested.
  • The legal entity, titles and a domain mailbox are owner gaps.
  • We cannot confirm that third-party profiles for sameAs exist until the owner supplies them.
  • Promised response times cannot be observed.

Visual quality and consistency

  • Only Chromium is rendered.
  • Only 1440 and 390 are captured, in the default theme.
  • Motion is judged from stills taken with animations disabled.
  • Aesthetic quality is subjective.

Start with the free scan

This free scan is a quick first look at your crawler settings and your homepage. Our team’s scan reads more of your site, fixes what it finds and re-checks it.

Scan my siteAsk for a walkthrough

Page last updated