Our method
Understand the evidence. Know the next step.
We examine a small sample of your public website and explain what needs attention. There are 15 checks in four sections, plus additional AI-reader settings. We give no overall or section grades.
Your free preview shows section summaries and up to three evidenced priority findings with broad next steps. The complete captured assessment is not included in that preview. Visit SplashCandy for the company and its services, or talk to SplashCandy about the support you need. You can use the preview with anyone you choose.
A quick snapshot of the starting page, up to two relevant public destinations and captured publishing rules. A desktop browser may run public-page JavaScript to capture visible text. We preserve the original HTTP responses and headers separately from that browser text. We do not sign in, submit forms or complete purchases.
How to read a result
Meets this check: captured evidence supports this specific requirement. Issue observed: we can point to evidence of a problem. Needs review: the evidence, its meaning or an attempted review remains unresolved. Not assessed: a substantive review was unavailable or had insufficient reliable input.
Eight checks use captured technical evidence. Seven use an automatic model review of captured browser text when the reviewer is configured and the input is sufficient. Those judgments cite exact captured quotations and follow the selected website purpose and visitor question. The report identifies whether review completed, was unavailable or failed; a failed review preserves the technical observations. Model judgments can be disputed and do not establish business truth. An unknown result is never counted as a pass or failure. Counts describe coverage, not the quality of the entire website.
One useful page can be sufficient. We do not require prices or additional pages where the purpose does not call for them. Partial renders, captured-text limits and shorter review excerpts are disclosed; information missing from a sample is not proved absent from the whole website.
Search foundations · checks 1–4
Search crawler permissions
What we measure: captured robots.txt rules for selected search crawlers and sampled paths.
Why it matters: these settings can restrict crawling of pages you want found.
What to fix: correct unintended restrictions on public pages. Keep deliberate restrictions. Permission does not prove crawling or indexing happened.
Page indexing instructions
What we measure: indexing exclusions in captured HTML and response headers.
Why it matters: noindex can exclude a page from search results even when it can be visited.
What to fix: remove unintended exclusions from pages you want indexed. Confirm their purpose first.
Canonical and page declarations
What we measure: malformed or conflicting preferred-page and identity declarations supplied in the sample.
Why it matters: inconsistent declarations give search systems conflicting guidance about the preferred page.
What to fix: correct proven invalid or conflicting declarations. A canonical pointing to another domain needs its intent checked. Missing optional markup is not a defect.
Page title matches its content · requires review
What we measure: whether the title accurately describes the actual subject or offer.
Why it matters: a descriptive title helps people understand which page they are choosing.
What to fix: read the page and replace vague or misleading titles with a concise description of its content. A title tag's presence alone cannot establish a good match.
Basis: Google's technical requirements and canonical guidance. These concern settings and eligibility, not guaranteed rankings.
Content clarity and useful information · checks 5–7
Purpose and audience · requires review
What we measure: whether a reader can understand the subject or offer and whom it serves.
Why it matters: visitors need to recognize whether the page addresses their needs.
What to fix: state the actual subject, service or product in concrete language and identify its audience. Short, useful explanations can be sufficient.
Information needed to decide · requires review
What we measure: whether the sample answers visitor questions relevant to the selected purpose.
Why it matters: missing scope, buying conditions or essential answers can prevent informed decisions.
What to fix: answer the specific unanswered questions. A service, shop and information page need different answers; universal pricing and word-count rules do not apply.
Statements agree in context · requires review
What we measure: conflicting statements about the same subject, location and time.
Why it matters: contradictions can confuse visitors about what is available or true.
What to fix: reconcile confirmed contradictions and date changes clearly. Different branches or nearby service areas are not automatically contradictory.
Basis: Google's helpful-content questions. They support reviewing purpose, completeness and clarity; they provide no automatic quality formula.
Identity and source information · checks 8–9
Provider or publisher identity · requires review
What we measure: whether captured content identifies the relevant provider, publisher or author and their role.
Why it matters: visitors need to understand whose offer or claims they are considering.
What to fix: make that identity clear. This does not verify legal identity or the truth of business claims.
Relevant contact or source context · requires review
What we measure: contact, buying or source-attribution information appropriate to the page's purpose.
Why it matters: visitors need suitable follow-up information, buying conditions or source context.
What to fix: supply the relevant information itself. An About, Contact or Terms link alone does not establish that its destination contains the answer.
Provider identification records a claimed source. A named business clearly speaking as “we” can be self-attribution. This does not verify its real-world identity, make its offer clear or establish sufficient contact information; those are separate questions. A logo, unrelated name or unnamed “we” alone does not identify who is responsible.
Basis: source and “Who” questions in Google's helpful-content guidance. Identity cues are information to inspect, not proof of trustworthiness.
Public routes and technical basics · checks 10–15
Sampled address responses
What we measure: the HTTP response at each selected public address.
Why it matters: a confirmed missing page or server error prevents access at that address.
What to fix: restore the intended page, correct its link or redirect the old address as appropriate. Denied or failed collection remains uncertain.
Readable content in returned HTML
What we measure: observable text in returned HTML, instead of an empty or unverified shell.
Why it matters: some readers depend on content available without JavaScript.
What to fix: investigate missing source content and serve meaningful initial HTML where appropriate. This does not establish the rendered appearance or the text's usefulness.
Relevant next destination · requires review
What we measure: whether a relevant next step reaches the intended captured answer or destination.
Why it matters: visitors need a suitable route toward their purpose.
What to fix: correct misleading destinations and supply missing answers. Self-contained information may need no further destination. Form and checkout completion are not tested.
Link names in the source
What we measure: names supplied directly or through supported referenced labels in captured HTML.
Why it matters: labels help people understand navigation.
What to fix: provide meaningful names for confirmed unnamed links. Unsupported naming mechanisms need review. This is not a complete accessibility or keyboard test.
Mobile viewport instruction
What we measure: whether HTML supplies a usable viewport width instruction.
Why it matters: this tells mobile browsers how the page should scale.
What to fix: add or correct missing or invalid instructions, then inspect the phone layout separately. A correct instruction does not prove a usable mobile design.
Supplied structured-data integrity
What we measure: parsing errors and inspected-field conflicts in supplied machine-readable declarations, within parser limits.
Why it matters: broken or conflicting declarations can misdescribe the page to systems reading them.
What to fix: repair established syntax or field errors. Empty or untyped declarations need review; absent optional schema is not a failure. Valid syntax does not prove truth, rich results or AI citations.
Basis: W3C link-purpose guidance, Chrome's viewport explanation and Google's structured-data policies. These observations cover only part of those broader requirements.
Additional AI-reader settings
We name selected readers and show captured permissions. Search discovery, reading requested by a user, and model training have different purposes. Permission does not prove a provider visited, indexed, cited, recommended or trained on a page. Training choices are not website defects.
See Google's AI-feature guidance and OpenAI's crawler explanations. We do not run AI-query or citation benchmarks or require special AI files.
Scope and your choices
Every report names its sample, collection date, evidence types, review state and unresolved checks. Capturing desktop browser text does not assess visual design or mobile usability. We do not measure loading or interaction performance, complete accessibility, actual search rankings, AI-answer visibility, business truth, forms, inboxes or purchases. Findings do not predict revenue or guarantee rankings.
Download your temporary report and use it with anyone. Only an explicit consented help request sends your supplied details and report context to the configured destination. Missing configuration or a failed send is shown plainly.
Think we missed something? Request a free review of a disputed finding, with no purchase required. This corrects disputed findings; it is not a full free site investigation.
Earlier saved reports
Earlier versions retain their original labels and arithmetic. They are historical results and are not silently converted to this score-free method. New reports use the 15 explained checks above.