A Technical SEO Audit with Semalt: A Step-by-Step Guide for Romanian Websites

The technical audit is the part of SEO everyone postpones. Not because it is hard, but because it produces lists of 400 problems of which 380 do not matter. An automated report tells you that you have 1,200 pages without a meta description; it does not tell you that 1,150 of them are filter pages that should never be indexed anyway. The difference between a useful audit and a useless one is not the tool, it is the order in which you read the data.

The guide below is the workflow we use on Bucharest projects, run entirely inside the Semalt platform. It takes about two hours for a mid-sized site and ends with a document you can send straight to development.

Step 0: preparation, 10 minutes

Before any scan you need three things: access to the Search Console property, a list of the 20-30 pages that matter commercially, and one reference competitor. The last one is the most often forgotten, but without it you have no benchmark: a 2.8-second load time looks bad until you see the market leader sitting at 4.1.

Open the panel at semalt.com/authorize, add the domain and connect your Google account with read-only access for the Search Console side. Add the competitor to the rank tracking module as well - you will need it in step 4.

Step 1: what Google says it found

The first thing we read is not our own crawl but real data from Google. The reason is simple: a proprietary crawler shows you what a bot could find; Search Console shows you what Google actually found, what it indexed and what it refused.

In the Search Console analytics section, read in this order:

1. The number of pages receiving impressions versus the number of pages you think you have. If you have 3,000 products and only 900 pages received any impression in the last 90 days, you have a discovery or quality problem, not an on-page optimisation problem.
2. Device split. In Romania, mobile share on commercial queries frequently exceeds 70%. If your average mobile position is more than two positions worse than desktop, you have a technical problem on mobile, not a content problem.
3. Country split. Diaspora traffic behaves differently. Separate it now, otherwise every average that follows will mislead you.

Warning signs at this step

High impressions and low clicks on category pages means weak titles, not weak positions. Steady clicks on pages you never promoted means you have valuable content you are not exploiting. Both situations are fixed in hours, not months - which is exactly why we read them first.

Step 2: the automated site audit

The automated audit included in AutoSEO and FullSEO campaigns scans the site and produces on-site recommendations. This is where the three-bucket triage rule applies, recorded in a simple table:

The third bucket is what saves your relationship with the development team. Send 400 tickets and you get zero back. Send 12 tickets with clear reasoning and you get 12.

Step 3: indexing, the part most people skip

A site can be technically flawless and completely invisible if its pages are never discovered. On projects that publish frequently - a blog, a shop adding products weekly, a listings site - the gap between publishing and discovery is often two to three weeks.

In the fast indexing module you submit the full sitemap or only the new URLs and watch verified bot visits in the live log: GoogleBot, BingBot, OpenAIBot. The first 100 URLs are free, which is enough for an honest test.

The test we recommend: pick ten pages published in the last 30 days, submit five through the module and leave five to the natural rhythm. Compare after two weeks. If the difference is significant, you have found a real bottleneck in your publishing process - probably an infrequently regenerated sitemap or insufficient internal links to new content.

What a verified bot visit means

The log shows the actual request and the response code. That is the difference between "I submitted the URL" and "the bot arrived and got a 200 OK". If you see 200 OK but the page still does not appear in the index after two or three weeks, the problem is no longer discovery but perceived page quality - an entirely different chapter of work.

Step 4: comparison with competitors

This is where the Google rank tracking module comes in, which works on any domain with no ownership verification. You are not looking at positions in general, but at one thing: on queries where the competitor is in the TOP 3 and you are not in the TOP 10, what page type is ranking for them?

The patterns that show up most often on Romanian sites:

• Category versus product. You compete with product pages on generic category queries. The fix is a category page with real text, not an empty product list.
• Guide versus service page. On informational queries ("how do I choose", "how much does it cost") articles win, not sales pages. If you want those queries you need informational content that links internally to the commercial page.
• Local versus national. On "company X Bucharest" the pages that win carry explicit local context - address, area, opening hours - not the generic national page.

Step 5: visibility inside AI answers

A technical audit in 2026 does not end at classic Google. The AI visibility module shows whether language models cite your site, on which queries, which pages get cited and who appears instead of you.

From a technical perspective three concrete things matter. First: content that can be extracted from HTML without executing JavaScript has a far better chance of being cited. Second: pages that answer a question directly, with a clear paragraph under a question-shaped subheading, get cited more often than pages that dance around the answer. Third: structured data and clear context about who you are - address, author, date - help with attribution.

The check is free and works on any domain, so you can compare your site directly with the competitor who does appear in answers.

Step 6: the remediation document

The audit ends in a single document with at most 15 actions, each carrying four fields: what changes, where exactly (URL or template), why (which metric is expected to move) and how we verify after implementation. Without the fourth field, nobody will know a month later whether it worked.

An example of a complete line: "Add 300-400 words of category copy to the category template (all 42 categories). Reason: competitors rank with category pages on generic queries while we rank with product pages. Verification: position on the 12 tracked category queries, six weeks after publication."

Three site types, three different audits

The same workflow applied to three realities we meet constantly in Bucharest. The difference is not in the tools but in what you read first.

Online shop with many product variants

Here the number one problem is dilution: the same sofa in seven colours generates seven nearly identical pages, each with its own title. Google picks one, usually not the one you would have picked. What to check first: how many variants receive impressions (if only one, the rest are ballast), whether the category page has copy of its own, and whether filters generate indexable URLs. At a client with 4,000 SKUs, correctly blocking filter URLs cut crawled pages by 60% in three weeks, and commercial pages started being revisited twice as often.

Local services site

Five to ten pages, hard competition on immediate-intent queries. The technical audit is short, but two things matter disproportionately: mobile speed (most searches come from a phone, often on the move) and consistency of contact details across the site, the Google Business Profile and local directories. Here, an hour spent cleaning up address inconsistencies is worth more than a week of title optimisation.

Bilingual RO/EN B2B site

The most delicate case. Check, in order: whether every RO page has its EN counterpart declared reciprocally through hreflang, whether the canonical version is the right one for each language, and whether the two versions genuinely differ in text - a careless machine translation produces pages Google treats as duplicates. The clear warning sign: in ranking data, the EN version appears on Romanian queries or the other way round. It shows up within days if you track both query sets in parallel.

Internal links: the audit no tool does for you

Tools tell you how many internal links a page receives. They do not tell you whether those are the right links. The manual check takes 30 minutes and is worth doing in every audit.

Take the 20 important commercial pages and for each answer three questions: how many clicks from the homepage it takes to reach it, which anchor text it receives most often, and whether it is linked from the relevant informational content. The pattern we find almost every time: the blog articles that bring traffic never link to the page that sells. It is the cheapest growth available in a project - one contextual link from the highest-traffic article to the right commercial page, added in five minutes.

The practical rule: no revenue page should sit more than three clicks from the homepage, and anchor text has to be descriptive rather than "click here". Both are quick to check and fixable without a developer.

How often to repeat the audit

For a stable site, a full audit every six months is enough, with monthly checks on indexing and positions. For a site in active development, where new templates ship monthly, the technical check has to run after every major release - most of the serious problems we have seen did not appear on their own, they were introduced by a deploy.

How to hand the audit results to the team

A good technical audit most often fails at handover rather than at analysis. The document reaches a developer who does not do SEO, gets read halfway and sits in the backlog for three months.

What works in practice: for every action, write the concrete "it currently looks like this, it needs to look like this" example, with the relevant code snippet or screenshot. Group actions by template rather than by page - "all category pages", not 42 separate URLs. And estimate the effort together with the developer before prioritising, because a change that looks trivial to you may touch a component used across the whole site.

The last element, most often forgotten: agree from the start who verifies after implementation and when. Without that, half the fixes are implemented partially and nobody notices until the next audit, six months later.

Mistakes that ruin a good audit

Auditing before you know what matters commercially

Without the list of pages that bring revenue, any audit becomes an alphabetical list of problems. Prioritisation comes from the business, not from the tool.

Confusing lab speed with real experience

A lab score says nothing about the user in Berceni on a congested 4G connection. Field data in Search Console is more relevant than any synthetic score.

Fixing everything at once

If you change URL structure, title templates and internal navigation on the same day, you will never know what worked. Stage the work in two to three week intervals.

Frequently asked questions

Do I need a paid campaign for the audit?

For steps 1, 3, 4 and 5 - no, the analytics side and the indexing check cover everything. The automated on-site audit is part of the managed campaigns.

How long until fixes show results?

Indexing problems show within days. Page-type and content problems, in 4-8 weeks. Authority problems, in months.

Does it work for WordPress and Romanian ecommerce platforms?

Yes. The checks run against delivered HTML and Google data, so the platform behind the site does not matter.