By Gecko Studio · updated October 2026
Most of the SEO audits on the market are, in reality, a Screaming Frog crawl, three screenshots from Search Console and a PDF full of red, amber and green traffic lights. Two hours of work. An invoice of eight hundred euros. And a vague feeling in the client that they now "know" what is wrong with their website, when all they know is what the tool was able to look at on their behalf.
At Gecko Studio we have been carrying out SEO audits for years, and we stopped calling them "audits" in the usual sense a while ago. What we deliver has more than 60 review points spread over five phases, takes between three and five working days even with automated pipelines behind it, and the final deliverable is not a pretty PDF: it is a set of analyses, prioritised decisions and an action plan that the client can show their team without feeling foolish. This article sets out what we review, why we review each thing, and why almost no agency does it this way.
The five phases at a glance
- Phase 0 · 10 points · Context: market analysis, real 360° competition, buyer personas, initial content strategy.
- Phase 1 · 18 points · Deep technical work: cross-checked crawling, real on-page, lab and CrUX Core Web Vitals, mobile on a real device, hreflang, cross-referenced indexing.
- Phase 2 · 12 points · Architecture and content: real map, optimised proposal, keyword-to-URL mapping, semantic internal linking, content decay, editorial line.
- Phase 3 · 12 points · Signals most people ignore: E-E-A-T page by page, entity versus keyword, schema, backlink profile, real local presence.
- Phase 4 · 8 points · What almost nobody audits in 2026: visibility in ChatGPT, Perplexity, Claude and Gemini, AI bots, GEO (Generative Engine Optimization).
Why 60 points and not the usual 15
Most audits sold as such cover the same ground: title and meta description, loading speed, broken links, duplicate text, sitemap, robots.txt and a couple of other things. Between ten and fifteen checks. All valid. None sufficient.
The problem is that SEO in 2026 is not decided by those fifteen things. It is decided by the forty-five that nobody is reviewing because they do not appear in any tool. The real search intent of the ideal customer. The hierarchy of entities Google has learned about your brand. The consistency of your E-E-A-T signals over the years. The way AI models cite you or ignore you when somebody asks about your sector. The health of internal linking in semantic terms, not just in number of links.
Our methodology splits all of this into five clearly different phases. The first does not touch code or conventional SEO tools. The next ones drill progressively into technical detail. And the last is where we decide what to do with everything we have found. Let us go through them.
Phase 0: the context no tool will give you
Before opening a single crawler, before looking at a single keyword, before even entering the client's website with technical eyes, we carry out four blocks of analysis that set the course for the rest of the work. They are the most ignored by the market and, in our experience, the ones that change the final diagnosis most.
Market analysis. What is happening in the client's sector right now. Seasonality, trends, competitive moves, barriers to entry, the real size of the pie. Not to fill a slide deck: to understand whether the SEO strategy we were thinking of proposing makes sense given the size and maturity of the market. In some sectors, fighting for first position on a generic keyword costs more than coming second on five better-qualified long-tail terms, and that is only visible with the context in front of you.
360° competition. Not the SEO competitors (those that rank for your keywords) but the real business competitors, the indirect digital competitors, and the aggregators or marketplaces that take a slice of the pie. In the tourism sector, for example, the SEO competitor of a boat charter company is rarely another charter company: it is SamBoat, Click&Boat, Booking. If your audit does not identify that, you are playing a game that is not yours.
Real buyer personas. Not the avatars from the marketing textbook: people with a problem, a budget, a specific time of year, a device they search from and a level of knowledge of the sector that shapes how they word their searches. Without this, everything that follows will be skewed towards the obvious keywords that, in practice, only the business owner searches for.
Initial content strategy. What kind of content would need to exist to connect what the market is searching for with what this business can offer. This is not yet an editorial calendar: it is a strategic framework that then conditions how we read the site's current architecture and which gaps we spot.
Phase 1: the technical part you already know, but in depth
This is where 80% of what the market calls an "SEO audit" lives. It is right that it exists, but it is no more important than the other phases, only easier to sell because it fits best into spreadsheets. What we look at in this phase:
Technical and crawl errors. A full crawl of the site with several tools cross-checked, not just one, because each sees different things. We count response codes, chained redirects, canonical chains, loops, orphan pages, pages with intentional or accidental noindex, robots.txt, meta robots and HTTP headers. Not just the quantity: the pattern. Twenty 404s scattered across a site of ten thousand URLs are not the same as twenty 404s all concentrated in the commercial branch.
Real on-page. Titles, metas, H1 to H6 hierarchy, semantic density, duplicates, thin content, cannibalisation. Here we cross the tool with a human eye, because a ranker cannot tell the difference between a well-placed H2 and an H2 that exists because the parent topic is badly framed.
Speed and Core Web Vitals, both in lab data and in real data from the CrUX dataset. A critical difference: the lab tells you what could happen, the field tells you what is actually happening to your user. If a tool shows you only one of the two, it is half lying to you.
Real mobile. Not just "the site is responsive": how the site performs on the device your user has, over the connection your user has, in the language and settings your user uses. For tourism clients on an island, half the traffic can come from visitors on a phone abroad, on roaming 4G. Auditing from a MacBook on fibre does not tell you that.
Hreflang and multi-language if the client works with several markets. This is where most silent mistakes happen: pages that should be served in English but are served in Spanish to the wrong user, broken hreflang implementations that nobody notices because they still do not produce a visible error in Search Console.
Real indexing, cross-referenced between Search Console, server logs where available, the sitemap and the crawl. And, most importantly: how many of the client's pages are indexed but receive zero impressions. Those pages are one of the clearest indicators of deep SEO health and almost nobody looks at them.
Phase 2: architecture and content, where the game is played
A site's information architecture decides whether its SEO is going to scale or not. A bad architecture can cover up any decent technical work. A good one forgives technical mistakes and mediocre keywords.
Map of the current architecture. Not the menu: the real architecture as Google sees it. Which pages carry weight, which are orphaned, which hubs exist and which should exist. How topics are grouped. What inconsistencies there are between the user's menu and the real graph of internal links.
Optimised architecture proposal. With this we already answer what we would change, with what priority and what impact we expect. Sometimes the answer is "don't touch the architecture, it is fine, the problem is elsewhere". That "don't touch" is also a finding of the audit, and one of the most valuable, because it avoids spending budget on refactors that add nothing.
Keyword-to-URL mapping. Which keyword each URL should rank for, which one it actually ranks for, and where there is serious cannibalisation dragging down the site's performance without anybody seeing it.
Semantic internal linking. Not the link count, which is what any tool gives you: the distribution of internal authority towards the pages that matter to the business. Many sites have all their internal link equity concentrated on "About us" and none on the pages that bring in revenue.
Content analysis, not only whether it exists but whether it works. Outdated content, decaying content, content that should be updated, content that should be merged, content that should be deleted. Content decay is one of the audits that generates most immediate impact and one that few agencies carry out, because it requires crossing Search Console history with the crawl, and that does not come out of a single tool.
Blog and current editorial strategy. Is there an editorial line? Is publishing frequent? Is there an identifiable content pillar? What role does the blog play within the commercial strategy? A blog that publishes two articles a month with no common thread is not content, it is expense.
Phase 3: the signals most people ignore
This is where most audits start to lose their pulse. The following checks do not appear in any automatic tool, give no pretty metrics for a presentation, and require the kind of work that consumes real time from a person who knows what they are doing.
E-E-A-T signals audited one by one. Experience, Expertise, Authoritativeness, Trustworthiness. For each important commercial page: who signs the content? Is there a visible author, with a bio and a public track record? Are there real cases, first-hand data, real measurements from the business? Are there visible trust signals off the site (reviews, press mentions, presence in serious directories)? Does what the client says about themselves match what the rest of the internet says about them?
Entity versus keyword. How the business is represented in Google's knowledge graph, which entities Google's Natural Language API recognises in the main content, how coherent the brand narrative is across the site, and whether there are contradictory entity signals confusing the ranker.
Schema and structured data. Not just "does it have schema?" but "does it have the right schema, properly nested, with complete properties and consistent with the visible content?". Most implementations we see are copied from generic examples and carry semantic errors that Google ignores, but which stop the site appearing in rich snippets.
Backlink profile with a critical eye. Not the Ahrefs DR score: the analysis of whether the profile is defensible, whether there are toxic patterns, whether the anchor diversity is natural, and above all whether the links reflect the reality of the business or are accumulated noise from the last few years.
Real local presence. Google Business Profile reviewed point by point, NAP consistency cross-checked against local directories, reviews, categories, photos and how fresh they are. For tourism clients this is usually the biggest bottleneck in local SEO and the least audited.
Phase 4: what almost nobody audits in 2026
In the past eighteen months, three new layers of visibility have appeared that already condition how much real traffic a business gets, and that practically no audit on the market covers. We put them into our methodology as soon as we saw them working with real clients.
Visibility in AI models. When someone asks ChatGPT, Perplexity, Claude or Gemini about your sector in your area, are you mentioned? In what context? Which sources is the model citing to build that answer? We audit visibility in the four main answer engines, measure share of voice against competitors, and identify which content is being cited and which content could be if it were rewritten with a different approach.
AI bots. Which AI bots are visiting the client's website, how often, which pages, and what strategic decision has been taken about it. Do we block? Let everything through? Open some sections and close others? We audit an extended robots.txt and the structured markup of the content. What we do not sell is an llms.txt file: Google says no special AI files are needed to appear in generative search, and we explain why here.
GEO: Generative Engine Optimization. How ready the client's content is to be consumed by a model, cited by an agent and returned as an answer in a conversational search. Structure, clear definitions, verifiable data, citations of primary sources, clear authorship. Without this, content can be perfectly optimised for Google and invisible to ChatGPT.
How we prioritise 60 findings into a plan that fits three months
If you carry out a 60-point audit, you end up with a list of between 40 and 150 findings depending on how bad the site is. Delivering that list as it stands is the most common mistake on the market and the one that frustrates clients most: nobody can carry out 150 things at once, nobody knows which one moves the needle most, and the report ends up in a drawer.
Our prioritisation process crosses four axes for each finding: expected impact on traffic and conversions, real implementation cost (not the theoretical one), the technical risk that the fix breaks something, and contextual urgency. What comes out is an action plan ordered into two-week sprints with clear owners, executable with the hands and budget the client has right now, not with ideals.
The final deliverable is not a PDF: it is a two-page executive summary for management, free of jargon; a detailed technical report for the team that will carry out the work; an action plan with dates; and a live tracking dashboard the client can open at any time to see where we are. Each of these speaks the language of whoever will read it, and none is padding.
Why almost nobody does this: the honest answer
We do it because we can. Most agencies cannot, and it is not for lack of will. There are three real reasons.
The first is economic. A well-done 60-point audit, with all five phases, consumes between 25 and 40 real hours of a senior professional. At the price the market has trained clients to expect (SEO audit = 500 to 1,000 euros), that is simply unsustainable. The only way out for an agency with that pricing is to cut phases, automate what should not be automated, and sell as an audit what is really a superficial report. We charge what we charge because the work that counts costs what it costs, and our proposal for each client explains exactly where each hour goes.
The second is capability. Part of what we look at is not learned by reading an SEO blog: it requires years of working with real clients, getting things wrong, and seeing what works and what does not. You cannot hire a junior and ask them to audit entity signals, because they have not seen enough cases to recognise the patterns. And training a person to get there takes years and loyalty: two scarce things in this industry.
The third is tooling. To do this in reasonable time we have built our own stack of pipelines that automates 60% of the repetitive work: data cross-checks, pattern detection, generation of intermediate reports. That frees the senior person to spend their time on what a machine cannot see. The stack is our own, we have been developing it for years, and it does not exist as a product you can buy anywhere. Without it, our own audit would cost us twice the time and be hard to sustain.
How to tell whether the SEO audit you are being sold is serious
If you are considering buying an SEO audit from any agency, including ours, these are the questions you should ask and the answers you should expect.
How many real hours will a senior professional spend on my case? If the answer is under fifteen, you are buying an automated report with a nice name. If it is a reasonable range (between twenty and forty for an SME with a medium-sized site), at least there is work behind it.
What analysis is done before touching the technical side of the site? If the answer starts straight away with "we'll run a Screaming Frog crawl", the whole of phase 0 is missing. If there is market analysis, real competitors, personas and a preliminary editorial strategy, at least the approach is the right one.
How will you audit visibility in AI models? If they look at you strangely, it is 2026 and they are fifteen months behind. If they have a clear process (tests across several models, share-of-voice measurement, citation analysis), they are up to date.
Does the deliverable include a prioritised action plan or just a list of findings? A list of 80 unprioritised items is not a plan: it is a data dump. If a dump is what you are handed, the audit is of no use beyond the feeling of having done your homework.
Can I see a previous real audit, even with the names hidden? Any serious agency has examples it can show. If they cannot show anything, they probably have nothing to show.
A 60-point SEO audit is not a luxury or a technical whim: it is the minimum level needed to understand why a website is not delivering what it could, in a sector where the ranker is no longer just Google and where AI models are starting to decide whom to recommend before the user has clicked on anything. If you have spent two or three years investing in SEO without the results you expected, the problem is almost never in the 15 points a tool reviews. It is in the other 45 that nobody has ever reviewed. For a first look, what a website audit is is a good starting point, and you can get in touch any time.
