SEO audit: the complete checklist to find out what's holding your website back

A company that makes components for agricultural machinery rebuilt its website in the spring. In September the marketing manager opens Search Console and sees that clicks from organic search have dropped by a third compared with the year before. They try a free analysis tool: 412 issues, 37 of them “critical”. Images without alt text, titles that are too long, a few broken links. They fix almost all of them in two weeks. The clicks don’t come back.

There was only one reason: during the rebuild, the old product pages hadn’t been redirected to the new ones. The other 411 issues were real, but they had nothing to do with it.

The short answer: an SEO audit is an orderly check of a website to understand what is stopping its pages from being found and in what order to act. You always start with indexing and crawling, because a page Google can’t see can’t rank, then architecture, content, on-page, speed, structured data, backlinks, local search and AI answers. Search Console, Screaming Frog, SEOZoom, Ahrefs and PageSpeed Insights cover almost every check. The result that matters isn’t the list of errors, but a list of fixes ranked by impact and effort.

If the problem concerns a single page, it’s better to start from the 24 reasons why a page doesn’t rank. Here we look at the whole website, following the sequence we use in our own audits.

What is an SEO audit and what should it deliver?

An SEO audit is an analysis of a website that lines up the technical, content and authority problems limiting its visibility in search results, and turns them into a work plan. An automated tool finds anomalies; the audit decides which ones matter.

At the end you should have three things in hand:

1. A baseline snapshot. Indexed pages, clicks and impressions, ranking keywords, Core Web Vitals status, link profile. You need it to measure afterwards whether the fixes worked.
2. The problems, with their cause. Not “312 duplicate titles”, but “the catalogue filter pages have duplicate titles because the theme builds them from the category name”.
3. The priorities. What to do this week, what this quarter, what can wait, and who has to do it.

Which tools do you need for an SEO audit?

A few are enough, as long as you know what to ask of each one.

Tool	What we use it for in the audit	Cost
Google Search Console	Indexing, clicks and impressions, Core Web Vitals, structured data, manual actions	Free
Screaming Frog SEO Spider	Response codes, redirects, canonicals, titles, page depth, internal links	Free up to 500 URLs, then an annual licence
SEOZoom	Ranking keywords in the Italian market, pages competing with each other, comparison with competitors	Paid
Ahrefs	Backlinks, domains citing competitors but not you, technical audit	Paid; free version for verified sites
PageSpeed Insights	Speed and stability of individual pages, with causes	Free

The free version of Screaming Frog stops at 500 URLs: enough for a brochure site, not for an e-commerce store. Ahrefs offers anyone who verifies their own site a free version with a technical audit and backlink data, with limits on the pages crawled each month.

Search Console is the only tool that reports Google’s own data on your site; the others make estimates. When the numbers don’t match, Search Console is what counts for your site. For competitors, whom you can’t see there, you need the others.

Where to start: indexing and crawling

Here, because every other check assumes Google sees the right pages. If an important page isn’t indexed, improving its copy won’t help.

The starting point is Search Console’s Page indexing report, which gives the reason for each excluded page. Some reasons are normal, such as a page with a redirect or a variant with a correct canonical. The ones we look at first:

• “Crawled - currently not indexed” on pages that ought to be there: often the content is judged too similar to something else or not very useful.
• “Discovered - currently not indexed” in large numbers: Google knows the URLs but hasn’t crawled them yet.
• “Duplicate, Google chose different canonical than user”: the canonical says one thing, the rest of the site another.
• “Excluded by ‘noindex’ tag” or “Blocked by robots.txt” on pages that should rank. This often happens when moving from the staging site to the live one.

Then you cross-check with a Screaming Frog crawl, which follows the site’s links and finds 404 and 5xx errors, redirect chains and canonicals pointing to pages that don’t exist. Comparing the crawl with the sitemap brings out orphan pages (in the sitemap but not reachable through links) and pages that shouldn’t exist, such as URLs with sorting or internal search parameters.

Has the site been rebuilt recently?

Then this check comes before everything else. Take the list of URLs from the old site, for example the pages with clicks in Search Console over the last 16 months, and check that each one has a permanent redirect to the most similar new page. Sending all the old pages to the home page is a common mistake: Google often treats it as a page that no longer exists.

Does crawl budget concern you too?

Probably not. Google’s guide to crawl budget is aimed at sites with more than a million pages that change every week, more than 10,000 pages that change every day, or a lot of URLs marked “Discovered - currently not indexed”. For an e-commerce store with thousands of filter combinations, though, the issue is real: we cover it in the article on SEO for e-commerce.

Are the important pages easy to reach?

Architecture tells Google which pages matter. A page linked from the menu and from twenty articles carries more weight than one reachable only from page five of the archive.

With Screaming Frog we check:

• Click depth. Pages that bring in leads or sales should be within three clicks of the home page.
• Incoming internal links. If the main service page receives three links and “about us” two hundred, the hierarchy is upside down.
• Anchor text. “Click here” says nothing; “grain drying systems” does.
• Orphan pages and internal links pointing to redirects or errors, to be fixed at source.

Then there’s a check no crawler does: does the structure reflect the way customers search? If you sell by application sector but the site is organised by product code, the page answering “valves for food processing plants” doesn’t exist.

Does every page answer a specific search?

This is the part that requires the most judgement and the one that usually makes the biggest difference. In its guide to creating helpful, reliable, people-first content, Google asks, among other things, whether the content provides original information or analysis and whether it was written or reviewed by someone who knows the subject.

We carry out four checks.

Intent. For the main keywords we look at what Google shows today. If the first page only has product pages with prices and demos, an informational article will struggle, and vice versa. The method is in the article on search intent.

Cannibalisation. Two pages competing for the same search, with Google alternating between them. In Search Console you can see it by filtering the Performance report for a query and opening the Pages tab; with SEOZoom you can check it at scale. It’s usually solved by merging the pages or giving them different purposes.

Thin pages. Product pages with only the supplier’s description, city pages with the same text, blog tags with a single article.

Outdated content. Prices, regulations or versions that are no longer current: a sign of little care for readers and for anyone deciding whether to cite you.

In the audit we don’t rewrite the copy: we indicate which pages to update, merge, delete or create, and why.

Do titles, descriptions and headings say the right thing?

These are the elements automated reports flag most, and they need to be read calmly. A 64-character title instead of 60 isn’t a problem; the same title on 300 pages is.

With Screaming Frog we extract the title, meta description and headings of every indexable page and look for:

• Missing, duplicate or generic titles (“Home”, “Products”). Google uses them to build the title link in search results, even though it can rewrite them.
• Missing or duplicate meta descriptions. They don’t move rankings, but if they’re missing Google chooses the text to show on its own.
• Missing or multiple H1s, for example the logo marked up as an H1 on every page.
• Images without alt text where the image carries information: product photos, diagrams, charts.

Then we cross-check with Search Console: pages with lots of impressions and a low CTR for their position are the ones where rewriting the title and description can bring in clicks without changing the ranking. It’s often the fix with the best ratio of impact to effort.

Is the site fast for the people who actually use it?

Speed and stability are measured with Core Web Vitals: LCP for loading, INP for responsiveness, CLS for layout shifts. What they measure and how to improve them is in our guide to Core Web Vitals; in the audit we’re interested in where the problems are and how much they weigh.

1. Search Console, Core Web Vitals report. It groups similar pages and tells you whether the problem is widespread (all product pages) or isolated.
2. PageSpeed Insights on sample pages from each group. It shows real data from Chrome users, which according to the PageSpeed Insights documentation covers the last 28 days, plus a simulated test with the likely causes.
3. Screaming Frog connected to PageSpeed, to collect metrics across hundreds of pages on large sites.

The simulated test and real-world data often don’t match, and for Google it’s the real data that counts. Low-traffic sites may not have any: then you rely on the test, knowing it’s an estimate. On their weight in rankings it’s worth being honest: Core Web Vitals matter, but they don’t outweigh the relevance of the content. We put them high on the list when they’re poor on pages that bring in revenue, because there they also affect conversions.

Is the structured data correct and still useful?

Structured data is code in the page that describes its contents to search engines: a product with price and availability, a business with address and opening hours, an article with author and date. It’s used for rich results and helps automated systems, including those that generate answers, understand who you are.

We check errors in Search Console’s reports and with the Rich Results Test; consistency between code and page, because prices and availability must be the ones the user sees; and missing types, such as Organization on the home page, Product on product pages, LocalBusiness for anyone with premises open to the public.

In the last two years another check has been added: making sure the types you use still have a visible effect. In 2025 Google announced the end of support for several little-used structured data types, and according to the documentation update log FAQ rich results no longer appear in Search from 7 May 2026. FAQ markup can stay, because it still describes the page, but it no longer earns extra space in the results.

Who talks about you, and how?

Backlinks, links from other websites to yours, remain one of the signals search engines use to judge how authoritative you are on a topic. In the audit we don’t count them: we look at where they come from and what’s missing.

With Ahrefs, and with SEOZoom for the Italian market, we analyse:

• Referring domains. Twenty links from trade magazines are worth more than two hundred from generic directories.
• Linked pages. Often almost all the links go to the home page and the service pages receive none.
• Links to pages that no longer exist. They’re common after a rebuild, and a redirect recovers them.
• The gap with competitors. Domains that link to two or three competitors but not to you are the starting list for digital PR.

According to Google, the disavow links tool should only be used if there are a lot of artificial or spammy links and they have caused, or are likely to cause, a manual action. Most sites don’t need it. If an audit suggests disavowing hundreds of domains without a manual action in Search Console, ask why.

Can people nearby find you too?

For anyone with a shop or practice, or serving a specific area, the audit includes local search. Google names three factors for local ranking: relevance, distance and prominence, meaning how well known the business is, including through reviews and links.

We check the Google Business Profile (correct primary category, opening hours, services, recent photos), consistency of name, address and phone number across website, profile and directories, the number and freshness of reviews and the replies to them, and the website pages dedicated to locations or areas served, which must have their own content rather than the same text with the town name swapped.

Do you appear in AI Overviews and ChatGPT answers?

Some searches now get a generated answer, from AI Overviews and AI Mode on Google or from ChatGPT, Perplexity and Copilot. The question becomes whether your site is among the sources cited.

Much of the work overlaps with the previous checks: for AI Overviews and AI Mode, indexing, snippet eligibility and content quality all apply. We add three checks:

• Crawlers allowed. Robots.txt and the firewall must not block the bots of answer engines, such as OAI-SearchBot for ChatGPT or PerplexityBot.
• Citable pages. The main pages answer the question in the first few lines, with passages that stand on their own.
• Measurement. Search Console’s report on generative AI features, visits from chatgpt.com and perplexity.ai in GA4, tools such as Ahrefs’ Brand Radar to compare the brand’s presence in answers with that of competitors.

The full list, engine by engine, is in our SEO checklist for AI search.

The SEO audit checklist in one table

The condensed version, with the main tool and the signal that makes us dig deeper.

Area	Check	Tool	Warning sign
Indexing	Excluded pages and reasons	Search Console	Important pages on noindex, blocked or “crawled - currently not indexed”
Indexing	Responses, redirects, canonicals	Screaming Frog	Linked 404 and 5xx errors, redirect chains, canonical chosen by Google different from yours
Architecture	Depth and internal links	Screaming Frog	Revenue-generating pages more than three clicks deep or with few internal links
Content	Intent and cannibalisation	SEOZoom + Search Console	Page type different from the first page of results; two pages alternating on a query
On-page	Title, description, H1, CTR	Screaming Frog + Search Console	Duplicate or generic titles; lots of impressions and few clicks
Speed	Core Web Vitals by group	Search Console + PageSpeed Insights	“Poor” groups on mobile
Structured data	Errors and consistency	Search Console + Rich Results Test	Errors on Product, different prices in code and on the page
Backlinks	Sources, linked pages, manual actions	Ahrefs + SEOZoom + Search Console	Links only to the home page, links to error pages, active manual action
Local	Profile and reviews	Google Business Profile	Wrong category, details different from the website
AI answers	Crawlers, citable pages, presence	Robots.txt, Search Console, GA4, Ahrefs	Answer engine bots blocked, no visits from chatgpt.com

How do you decide what to fix first?

With an impact-versus-effort matrix. Each problem gets two scores from 1 to 3: how much it could change traffic, leads or sales, and how much work it takes.

	Low effort	High effort
High impact	Do it now	Plan it for the quarter
Low impact	Do it in a batch, when the chance comes	Leave it

For impact we ask ourselves three questions. Does it affect pages that bring in leads or sales? Does it concern one page or a whole template, such as all product pages? Does it block something upstream, such as indexing, or worsen a detail? For effort, what matters is who has to do the work: whoever manages the content can rewrite ten titles in an afternoon, while changing how the theme generates filter pages requires a developer and a release.

An example, with figures made up for illustration: a homeware e-commerce store with 4,000 products.

Problem	Impact	Effort	Decision
180 old product URLs without redirects, with external links	3	1	Now
1,200 filter URLs indexed, almost identical to the categories	3	2	This quarter: noindex or canonical, out of the sitemap
Category titles that are just the name (“Cushions”)	2	1	Now
Poor LCP on mobile for all product pages	2	3	This quarter, with the developer
2,300 images without alt text	1	2	In a batch, starting with product photos
Meta descriptions over 160 characters	1	1	When the pages are reviewed

The first three rows are worth more than everything else. A report that puts them on the same level as long descriptions doesn’t help anyone decide.

How often should an SEO audit be repeated?

A full audit once a year for a stable site, and always before and after a rebuild or migration, after a traffic drop that seasonality doesn’t explain, and before starting ongoing SEO work. In between, monthly checks on indexing and new errors in Search Console are enough. The drop in the opening example would have shown up within a week with a check after go-live.

Mistakes to avoid

Treating the tool’s report as an audit. The tool assigns standard severity levels. It doesn’t know which pages bring in revenue or what changed on the site last month.

Starting with on-page details. Titles and alt text are easy to fix, but if the pages aren’t indexed or the redirects are missing, they change nothing.

Relying only on external tools. For your own site the reference data is Search Console’s. SEOZoom and Ahrefs are for the market and competitors.

Not saving the baseline snapshot. Without clicks, impressions and indexed pages recorded beforehand, after three months you won’t be able to say what worked.

Delivering the list without owners. Every fix needs someone to do it and a date. Otherwise the audit stays a PDF.

Want to know what’s holding your site back?

The audit is the first step in our SEO work: we analyse indexing, architecture, content, speed, structured data, links and presence in AI answers, and deliver the problems with their causes and priorities ranked by impact. We walk you through it on a call, because a list without explanations is no use. If you’re mainly interested in being cited by AI Overviews and ChatGPT, see how we work on GEO.

You can book a call with Federico or write to us via the contact form.

Frequently asked questions

How long does an SEO audit take?

It depends on the size and history of the site. For a few dozen pages, a few days’ work is enough; an e-commerce store with thousands of products, or a site that has just been migrated, takes longer, because the crawls are long and the problems need to be traced back to page templates.

Can I do an SEO audit myself with free tools?

Largely, yes. Search Console, PageSpeed Insights, Screaming Frog (free up to 500 URLs) and the free version of Ahrefs for verified sites cover indexing, on-page, speed and a first analysis of links. The hard part is deciding which problems matter and in what order.

What’s the difference between a technical SEO audit and a full one?

A technical audit checks whether search engines can crawl, index and show the pages: response codes, redirects, canonicals, sitemap, speed, structured data. A full audit adds content and intent, external links, local search and presence in AI answers.

How soon do you see the effects of an SEO audit?

Technical fixes, such as missing redirects or accidental noindex tags, can take effect within a few weeks, the time it takes Google to recrawl the pages. Content and links usually take a few months.

Does an SEO audit also help you get cited by ChatGPT and AI Overviews?

Yes, and most of the checks overlap. For AI Overviews and AI Mode, indexing, snippet eligibility and content quality apply; for the other engines you add a check on their crawlers and the measurement of citations and visits.