How to use BoltAudit.

Install it, run your first audit, apply a fix, undo it if you like. Two minutes to learn, and below it, the exact math behind every number we show you.

Four steps from install to fixed.

Step 1 of 4. Install the free plugin.

From WordPress.org. No account, no card, read-only by default.

Step 2 of 4. Run your first audit.

The Local Audit scans on your own server and names what is slowing you down.

Step 3 of 4. Apply a fix.

Ranked by what it wins back. You see the evidence and exactly what will change before you click.

Step 4 of 4. Undo anything.

Anything applied while one-click fixes were available is in Fix History, and can still be reversed there.

Paid plans add AI-ranked fixes, and the SEO & AEO audit arrives in the next version. See pricing

How it works: where our numbers come from.

Both of BoltAudit's numbers are estimates, and you deserve to see the working. Here it is, in plain English.

How does BoltAudit estimate visitor loss?

BoltAudit measures your site's real loading speed across phones and computers, then uses published research on how many people abandon a page at that speed to estimate the share of visitors you are losing. It weights the estimate by your actual mobile and desktop traffic split. It is a research-backed estimate, not a measurement of your analytics, and every figure is labeled as an estimate.

Measured speed in, an estimate out.

Step 1 of 4. We measure how fast your pages actually load.

Real speed on phones and computers, measured separately.

Step 2 of 4. We look up how many people give up at that speed.

Published abandonment research, not our opinion.

Step 3 of 4. We match it to your real audience.

Weighted by your actual mobile and desktop split.

Step 4 of 4. We rank every fix by how many visitors it wins back.

That ranked list is your Audit Report.

A worked example.

Preview

4.1s mobile load about 4 in 10 abandon at that speed 68% of this site's traffic is on phones estimated 24% of visitors lost

Biggest cause on this sample site: 4.2 MB of oversized images. Fixing it wins back an estimated 9%.

Where does the visibility score come from?

The score is an estimated composite of two things we can genuinely measure. Search readiness: deterministic checks on your own pages, titles, metas, schema validity, canonicals, internal links. AI citability: how ready your pages are for answer engines, crawler access, content structure, whether the answer sits in the HTML where an AI can read it. Each check carries a weight, each finding carries evidence, and when we cannot produce a defensible number you will see "estimate unavailable" instead of a guess.

Estimated Preview
Search readiness68
AI citability47
Visibility score61

What the confidence figure means.

Every finding carries a confidence percentage. It is not a probability that the finding is real, and it is not measurement error. Your site reported a number and that number is not in doubt. What the percentage scores is the judgement wrapped around it: whether the band the value falls in is the right band, and whether the threshold defining that band is well grounded. Three things go into it.

How the check decides. A check that asks whether your site has a thing — a page cache, a canonical tag — has no line to draw, so there is little to be unsure about. A check that measures a number has to pick where "too much" starts, and reasonable engineers put that line in different places.
How far past the line your site sits. A value a hair over the threshold would be under a slightly different but equally defensible threshold. A value several times over it is past every threshold anyone proposes. This is why two sites can trip the same check and see different confidence.
Where the threshold came from. Every check records its source. A line published by a standards body, by a vendor documenting its own product, or by WordPress itself is better grounded than one aggregated from community practice. That is not a dismissal of practice — several of these thresholds are best known that way — but it is a real difference and the score reflects it.

Confidence never reads 100%, because a threshold is a judgement and we do not claim certainty about one. It never reads below 50% either: if we expected a finding to be wrong that often, the honest move would be to remove the check, not to show you the finding quietly. Every finding also shows the sentence behind its own number, so you never have to take the percentage on its own.

What these numbers are not.

They are estimates, not your sales figures. We do not read your analytics.
The visibility score is not a ranking tracker, and it does not count your citations in AI answers. Nobody can measure those reliably today.
Read both numbers as direction and size, not decimals.
When we cannot defend a number, we say "estimate unavailable" instead of inventing one.

Then we measure whether it worked.

After you apply a fix, Fix History shows the real measured change, including "no measurable change". The estimate got you to act. The measurement tells the truth. That is also why the measurement, not the estimate, is what BoltAudit reports back.

Fix History Preview
Compress 41 images measured +8% mobile Applied
Defer offscreen scripts Rolled back by you Rolled back

Sources

Where research is estimated or contested, our figures say estimated.

Try it on your own site.

The Local Audit is free, read-only, and runs on your server. Two minutes to your first report.

Install the free plugin See SEO & AEO