How FixSift reaches its conclusions

What FixSift analyses

FixSift reads your website the way a first-time visitor does — the words on the page, the structure around them, and the markup underneath. It starts with your homepage, because that is where most first visits land, and then looks at a limited number of the pages a visitor would reach next.

Every scan covers seven areas: messaging, conversion, trust, user experience, SEO, AI Search Readiness and accessibility.

Structured checks, AI-assisted judgement

Some of what FixSift checks is objectively true or false: whether a page has a meta description, whether images carry text alternatives, whether headings run in order. Those checks are deterministic. They do not involve a model, and they produce the same answer every time.

The rest requires judgement: whether a headline explains what you do, whether the proof on the page would convince a sceptical buyer. FixSift uses AI for those, structured around the page's own content — and then validates every result against the facts extracted from the page before it reaches your report.

The two are not mixed. A deterministic finding is never overridden by a judgement, and a judgement that contradicts something extracted from your page is withdrawn rather than shown.

How the score is built

Each of the seven areas has its own score, built from the checks in that area that applied to your page. Your FixSift Score is a weighted average of the seven.

Every check resolves to pass, partial, fail, or not applicable. A check that does not apply to your kind of business is removed from the calculation entirely — it is not counted as a failure. A local plumber is not marked down for having no pricing table.

Scores are comparable between two scans of the same site as long as the scoring model has not changed in between. When it has, the comparison says so, and the report you already bought is never restated.

Evidence, and what FixSift will not do

Every finding points at something on your page. If FixSift says your headline is unclear, the report shows you the headline it read.

FixSift does not invent facts about your business. It will not claim you have customers you have not mentioned, prices you have not published, or certifications you do not display. Where a rewrite would be stronger with a fact FixSift cannot verify from your page, you get a clearly marked placeholder and a note about what to supply — not a plausible-sounding invention.

Findings FixSift is not confident about are withheld rather than hedged, and where the page gave too little to work with, no score is published at all.

What FixSift cannot tell you

FixSift analyses a website. It does not know your margins, your sales process, your market, or why a particular visitor did not buy. It cannot tell you whether a price is right.

It reads what a page serves to an automated visitor. Content that appears only after a login, only after an interaction, or only when scripts finish running may be missed — and the report tells you when that happened rather than scoring you on what it could not see.

Fixing everything FixSift finds will make your homepage clearer. It is not a guarantee of traffic, rankings, or revenue, and nobody who tells you otherwise is measuring the same thing.

AI Search Readiness, precisely

This area measures whether an AI assistant reading your page could describe your business correctly: is what you do stated in plain text, are your key facts machine-readable, does the page contradict itself, are the obvious questions answered.

It is not a ranking and not a placement prediction. No one can promise a position in an AI assistant's answer, and FixSift does not. What it measures is whether the raw material on your page is good enough to summarise accurately — which is the part you control.

Accessibility, precisely

FixSift checks the parts of accessibility that can be judged from markup: missing text alternatives, unlabelled form inputs, heading levels that skip, link text that says nothing on its own, declared page language, and obvious contrast problems.

This is not a WCAG conformance audit and it is not a compliance statement. Keyboard traps, focus order and real screen-reader behaviour can only be assessed by testing the live page, usually with a human and assistive technology. FixSift finds the machine-checkable problems and says so; it certifies nothing.