How We Run SEO Audits Today: Human Plus Pipeline

An audit run consists of four automated stages:

1. Crawl and technology: full site crawl, status codes, redirects, internal linking, indexability, hreflang. 2. Search data: Search Console queries across twelve months, trend detection, cannibalisation, declining pages. 3. Performance: Core Web Vitals from field data (CrUX) and lab measurement, per page type instead of per single page. 4. Structure and AI readability: schema validation, heading hierarchies, citability checks for AI search.

An AI pre-evaluation clusters the findings and proposes a prioritisation. Up to this point, minutes pass, not weeks.

What we need from the client for this is modest: access to Search Console, the domain, and half an hour of conversation about which pages carry the business. The pipeline fetches the rest itself.

The pipeline reliably finds everything. That is exactly its problem: a raw report contains hundreds of findings, from critical to cosmetic. Without prioritisation it is a mountain that paralyses teams.

My job has shifted accordingly. Instead of collecting data, I answer three questions: Which five findings actually cost this company visibility and inquiries? In what order is fixing them worthwhile, measured by effort and impact? And who implements what, agency or in-house team?

On top of that comes the interpretation no tool delivers: whether a traffic decline is a technical problem, a Google algorithm change or simply seasonality. That distinction has prevented more than one panic project.

So this does not stay abstract, an anonymised example from a real audit, typical for a mid-sized company website with two languages and a few hundred pages. The raw report of one run comprised 214 findings: hreflang errors on 40 pages, three redirect chains, twelve orphaned pages without internal links, plus dozens of missing alt texts and overly long page titles.

Three of them were business-relevant. First, the hreflang errors: the German and English pages referenced each other incorrectly, and Google partly showed the English version in German search results. Second, one of the three redirect chains, because it sat on the most visited service page and measurably cost loading time there. Third, two of the twelve orphaned pages, including, of all things, one that used to bring inquiries and had fallen out of the navigation in the last relaunch.

The remaining 211 findings were not wrong, but secondary. Exactly this separation is the consulting part: nobody would have recognised the orphaned pages as important without the cross-check against search data. And the alt texts would have kept an in-house team busy for weeks without bringing a single additional inquiry.

The nice thing about progress measurement: the follow-up run three months later showed the effect in black and white. The hreflang errors were gone, the German pages rank in German results again, and the reactivated orphaned page shows up in Search Console with clicks again. Three targeted corrections, measurably proven. That convinces a management board more than any 80-page report.

The honest answer first: for many teams, a bought SEO suite is the right choice. We still chose to build our own on the basis of open tools, for two reasons.

First, suites check what everyone checks. Our pipeline additionally checks what we consider relevant, such as the citability of content in AI search. We build such checks in days instead of hoping for a vendor's roadmap.

Second, insights from every audit flow back into the tools. If a finding type proves irrelevant three times, it gets removed from the report. A suite grows bigger with every release; our reports get shorter.

Where the suite wins, we say so too: competitive data and keyword databases. For those we continue to use external sources; the pipeline does not replace them. Building your own pays off for the part that carries your own consulting experience, not for data collections others maintain better.

An audit used to be a fixed-price project with a PDF at the end. Today we sell it as a first run plus repetition. The first run with a prioritisation workshop costs noticeably less than the old PDF project, because the data work disappears. Follow-up runs are cheaper again, because the pipeline is already set up for the website by then.

That shifts the budget to where it works: less money for the document, more for implementation and progress tracking. And it takes the pressure out of one-off-ness. An audit no longer has to solve everything at once, because the next run three months later measures what has happened anyway.

Internally, this also changed the sales conversation. We no longer talk about the page count of a report, but about the question of who implements the prioritised measures and by when. That is the conversation we always wanted to have; the PDF used to stand in the way.

That leaves the question of what arrives at the client's end. From the feedback of recent months, it is three things, in this order:

Speed: first solid findings after days instead of weeks. The prioritisation workshop happens while the numbers are fresh and the topic still has attention in the company.

Repeatability: the same run three months later shows in black and white what the measures achieved. An audit at arocom is no longer a snapshot but a progress measurement.

Honest budget allocation: consulting time flows into prioritisation and implementation instead of spreadsheet maintenance. For the same budget, you get more of the part that creates impact.

If you want to see what such a run looks like for your website: the GEO Score Check is the freely accessible mini version, the Future Check the complete one.