Local SEO Audit Tool for Google Business Profile

Your overall health score is a weighted average of four measured areas: reviews carry 40%, posts 25%, media 20% and services 15%. Nothing else enters the number.

Those weights are a claim about local SEO, and they are meant to be argued with. Reviews dominate because review volume, rating and reply rate are the signals most consistently associated with local pack placement — and the ones a business can actually influence this month. Posting carries a quarter because an inactive profile decays: gaps in activity show up in visibility well before they show up in any dashboard.

The thresholds are deliberately coarse. Above 80 the profile is in strong shape and the remaining work is detail. Between 50 and 80 there is a specific gap worth closing. Below 50 something fundamental is missing, and precision in the score matters less than fixing the obvious hole.

An earlier version of the score included a profile completeness component worth 20%. It was removed, and the weight was redistributed to reviews and posts.

The reason is risk. Rewarding profile edits pushes people to touch the fields that are most dangerous to touch: business name, primary category, address. Google re-verifies aggressively after such changes, and a suspended or re-verification-pending listing costs far more visibility than an incomplete field ever did. A score that quietly nudged you toward those edits would be optimising your dashboard at the expense of your listing.

So the audit reports profile gaps as recommendations — it just refuses to pay you points for editing sensitive fields. If a field genuinely should change, change it deliberately, not to move a number.

The number itself is calculated arithmetically from your own data. AI is layered on top for sector-specific commentary, and it runs with a 12-second ceiling: if the model does not answer in time, the audit completes and reports the score without it.

This split matters for a practical reason. A score produced by a language model is not reproducible — run it twice on unchanged data and you can get two different answers, which makes "my score went up" meaningless. A deterministic core means that when the number moves, something in your profile actually moved.

It also means the audit still works when the AI provider is rate-limited or down. You lose the sector commentary, not the assessment.

Description scoring is where the audit stops being generic. Length is only part of it: 500 characters or more earns the full length component, 250 earns most of it, 100 earns a quarter, and anything shorter earns almost nothing.

The rest of the score comes from cross-referencing. The audit takes the keywords you are actually tracking in the rank tracker and the service names defined on your profile, then checks whether any of them appear in your description. Keyword overlap and service overlap each contribute their own share.

This turns a vague instruction into a specific one. Instead of "add keywords to your description", the recommendation becomes: you are tracking these phrases, you offer these services, and neither appears in the text customers read. That gap is concrete and fixable in five minutes.

Service coverage is scored on its own count as well, in bands: eight or more services reads as complete, five is solid, three is thin, one is barely present. The bands exist because a service list is one of the few profile areas where more genuinely is better and the risk of editing is low.

Alongside the static score, the audit looks at a deliberately short recent window: new reviews in the last three days, at least two new media items, at least two posts.

Three days is short on purpose. A monthly view lets a profile look healthy while it has actually been dormant for weeks — the totals stay impressive and hide the stall. A three-day window catches the stall while it is still cheap to fix.

The distinction the audit draws here is useful: having no reviews at all is a different problem from having plenty of reviews and none this week. The first is a starting-point problem, the second is a maintenance problem, and they need different responses. Activity findings are reported separately from totals for exactly that reason.

Every recommendation is produced by a condition evaluated against your data, and its priority is computed rather than fixed. Review activity illustrates the pattern: no reviews at all is high priority, none in the last three days is also high, fewer than three is medium, and a healthy flow drops to low.

The consequence is that two businesses rarely get the same list, and the same business gets a different list next month. A finding disappears when the underlying condition stops being true — not when someone marks it done.

Each recommendation also carries the destination that fixes it, so the audit hands you the module rather than the homework. Reading a finding and acting on it are one step, not two.

Beyond the profile itself, the audit checks your presence across directories and compares what those listings say about you. Four fields are evaluated independently: business name, address, phone and website. Their agreement produces a consistency score.

Field-level checking matters because inconsistency is rarely total. A directory usually has your name right and your phone number three digits wrong, or an address from two moves ago. A single pass/fail verdict hides which field is actually broken; four separate checks tell you what to correct where.

Missing listings are surfaced as opportunities rather than errors. A directory you were never on is not damaging you — it is simply a place you are not yet found.

Because the weights are public, the arithmetic is predictable. Review volume and reply rate move the number most. Consistent posting is second. Media and services matter but saturate — once your service list is complete, adding a ninth entry changes nothing.

What does not move it: editing sensitive profile fields, since that component was deliberately removed. Re-running the audit without changing anything, since the core is deterministic. And keyword stuffing your description, since the check looks for presence of your tracked terms, not repetition of them.

Re-run cadence follows from this. Weekly makes sense while you are actively working on reviews and posting, because those are the components that move. Monthly is enough for maintenance. Running it twice in one day tells you nothing new.