Website SEO: What It Is and Where to Start in 2026

SEO is the work of making sure search engines can find, understand and show your pages, and that people choose them over other results. Google describes SEO the same way: helping search engines understand your content and helping users decide whether to visit your site.

Below: how Google Search works, the technical minimum, content requirements, the rules on spam and AI-written text, links, local SEO and a “what to check → where → target” checklist. Everything was checked against Google Search Central and web.dev documentation as of October 9, 2026.

What SEO means in plain terms

Google results include ads and regular, organic results. Advertisers pay for ads, but you can’t buy a spot in organic results: Google states plainly that it doesn’t accept payment to crawl a site more often or to rank it higher. SEO deals only with organic results.

There are no secrets that will automatically rank your site first — that’s Google’s own wording too. No one can guarantee the top spot, so a promise of “#1 in a month” is a red flag. Some changes show up in Search within a few hours, others take several months, and Google suggests waiting a few weeks before judging the results.

You don’t need separate “optimization for AI” either: AI Overviews and AI Mode run on the same ranking and quality systems, and Google calls the work of getting visible in them, including the trendy AEO and GEO labels, the same SEO.

How Google finds and shows pages

1. Crawling

   Googlebot finds the URL through a link or in a sitemap and downloads the page

2. Indexing

   Google analyzes the text, images and video, picks the canonical URL and stores the data in its index

3. Serving results

   For a query, Google searches the index for matching pages and picks the most relevant ones

Crawling. Google discovers new pages through links from pages it already knows and from lists of URLs (sitemaps) that site owners submit. Googlebot downloads the page and renders it in a recent version of Chrome, running any JavaScript. For indexing and ranking, Google uses the mobile version of a site, crawled with its smartphone crawler.

Indexing. Google analyzes the content, groups similar pages into clusters and picks a canonical page from each — that’s the one that may appear in results. Indexing isn’t guaranteed, even for a technically sound site.

Serving results. When someone enters a query, Google searches its index for matching pages and shows the most relevant ones. Results depend on things like the user’s language, location and device, so rankings may differ between countries and devices.

The technical minimum: getting your site indexed

Google Search Console

The free Search Console is a site owner’s main tool. Add your site as a “Domain property” (verified only with a DNS record at your registrar or host) or as a “URL-prefix property.” The first data will appear within a few days.

The URL Inspection tool shows whether a page is indexed and which canonical page Google selected. The “Request indexing” button sends a page to be recrawled, but the number of these requests per day is limited. The Page indexing report gives the overall picture: how many URLs are “Indexed,” how many are “Not indexed” and why.

Sitemaps

A sitemap is a list of the URLs you want to appear in Search. A single file holds up to 50,000 URLs or 50 MB uncompressed; if you have more, split the list into several files and combine them with a sitemap index file. Use full absolute URLs and canonical URLs only.

Google ignores the <priority> and <changefreq> tags and uses <lastmod> only if the date consistently matches when the page actually changed. You can submit the file in the Sitemaps report (the “Add a new sitemap” field → “Submit”) or with a Sitemap: line in robots.txt. It’s a hint, not a command: Google decides for itself what to crawl.

robots.txt and noindex

The robots.txt file sits at the root of the host (example.com/robots.txt) and manages crawling — mainly so crawlers don’t overload your server. It can’t remove a page from Google: a page blocked in robots.txt can still be indexed if other sites link to it, just without a description.

User-agent: *
Disallow: /admin/

Sitemap: https://example.com/sitemap.xml

To keep a page out of Search, use a noindex meta tag or an X-Robots-Tag HTTP header, and protect private content with a password. A page with noindex must not be blocked in robots.txt; otherwise the crawler won’t see the rule and the page may stay in results.

Canonical URLs

The same page often opens at several URLs: with and without www, over HTTP and HTTPS, with URL parameters. Google will pick a canonical version on its own, but you can give it hints. A redirect and a rel="canonical" tag are strong signals, including a URL in your sitemap is a weak one, and the signals stack. Don’t use robots.txt for this.

<link rel="canonical" href="https://example.com/seo-optimizaciya/">

For Google, a canonical URL is a hint, not a rule. To see which URL it actually chose, check the “Google-selected canonical” field in the URL Inspection tool.

Mobile version and HTTPS

Google recommends responsive design: one page adapts to any screen, and it’s the easiest pattern to maintain. If you have separate versions, the mobile one must have the same content, headings, structured data, title and meta description as the desktop one. Don’t load primary content only after a click or swipe.

Google’s page experience self-check includes serving pages over HTTPS and avoiding intrusive interstitials that cover the content. HTTPS also plays a role in choosing the canonical URL, so it’s best to move every version of the site to HTTPS with redirects from the old HTTP URLs.

Core Web Vitals: speed and stability

These are three metrics of real visitors’ experience. Google says its ranking systems use them, but there’s no single “page experience signal,” and when choosing between a pleasant page and a more relevant one, Search will show the more relevant one.

The target is measured at the 75th percentile of page loads, separately for mobile and desktop. INP replaced the older FID metric in March 2024. Real-user data is in Search Console’s Core Web Vitals report and in PageSpeed Insights; Lighthouse runs in a lab, can’t measure INP and shows Total Blocking Time instead.

The first part of LCP is the time to the first byte of the server’s response, so your hosting affects the result too (for a comparison of providers, see our ranking of web hosts in Ukraine).

Structured data

schema.org markup tells Google what’s on a page: an article, a product, a recipe, an event, a job posting. Google recommends the JSON-LD format and testing with the Rich Results Test. Markup doesn’t guarantee a rich result, and Google’s AI features in Search don’t require it.

The list of supported features changes. Google stopped showing FAQ rich results on May 7, 2026, and HowTo isn’t in the gallery either, so adding that markup to get “expandable questions” in results is pointless. For the current list, see Google’s structured data search gallery.

Content: what Google considers helpful

Google’s ranking systems aim to put helpful, reliable content made for people above content made to game rankings. Google suggests asking yourself: Does the content provide original information, research or analysis? Does it offer substantial value compared with other pages in search results? Is it free of easily verified errors? Will readers leave feeling they’ve learned enough to achieve their goal?

Warning signs of “search engine-first” content: writing to a specific word count because you’ve read that Google prefers one; publishing lots of content on many topics hoping something takes off; summarizing others without adding anything; changing the publish date without substantial edits. Content length on its own doesn’t affect ranking.

E-E-A-T: experience, expertise, authoritativeness, trustworthiness

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) describes qualities Google’s systems try to identify using a mix of signals. Trust is the most important. E-E-A-T isn’t a specific ranking factor on its own, but for YMYL topics (“Your Money or Your Life”: health, finances, safety) Google gives these qualities extra weight.

In practice: add bylines, tell readers about your authors and link to primary sources. Google explicitly calls fake author profiles — with AI-generated photos, made-up names or false credentials — deception. If a text was largely produced automatically, explain to readers how.

Title, description and headings

Google generates the title link in search results automatically. The main source is the <title> element, but Google also looks at the page’s visible main title, <h1> and other headings, og:title, large prominent text, anchor text of links to the page and WebSite structured data. To get Google to use your version, make the title unique, descriptive and concise, without repeated keywords or boilerplate. There’s no length limit, but the title link is truncated to fit the device width.

Google builds the snippet under the title mostly from the page content and uses the meta description when it describes the page more accurately. Write a short description with the key points for each page, not a list of keywords. Google doesn’t use the keywords meta tag at all. The order and number of h2–h6 headings don’t matter to Google, but readers rely on them to find their way.

Images and alt text

Google finds images in the src attribute of an <img> element; images set via CSS aren’t indexed. JPEG, PNG, WebP, AVIF and other common formats are supported. Place images near the text they relate to and give files short, descriptive names. Alt text should describe the image and how it relates to the content, not list keywords.

Internal links and URLs

Google follows only links that are <a> elements with an href attribute. Every important page should be linked from at least one other page on your site. Link text should explain where the link goes: “Core Web Vitals report,” not “here” or “read more.”

Use readable words in URLs rather than random IDs. Keywords in the domain name, though, have hardly any effect beyond appearing in breadcrumbs, Google says, so pick a domain for your brand (for how domains work, see our domain name explainer).

Links and site reputation

Links from other sites still matter: Google uses the PageRank algorithm but stresses that it’s only one of many factors. Natural links come when your site has something worth linking to and people hear about it — through social media, engaging with your audience, online and offline ads, and recommendations.

Buying links that pass ranking credit isn’t allowed — that’s link spam. Mark paid and advertising links with rel="sponsored" (Google prefers it but also accepts nofollow), and links from comments and forums with rel="ugc". Use nofollow when you don’t trust a site, not on every outbound link. If you sell placements yourself, the labeling rules are covered in our article on website monetization.

Google also counts as link spam excessive link exchanges, automated link building, low-quality directories, links spread across other sites’ footers, paid articles with links that pass ranking credit, and guest posts with keyword-rich anchor text. Domain “authority” scores from SEO tools are those tools’ own estimates: third-party tools have no access to Google’s internal ranking data.

What not to do: Google’s spam policies

Sites that violate the policies may rank lower or disappear from results. Google detects violations with automated systems and, when needed, human review. You’ll be notified of a manual action in the Manual actions report in Search Console; after fixing the problem, you select “Request Review” there. Since 2026, Google may also take manual action based on user spam reports, and the policies also apply to AI responses in Search.

• Cloaking — showing crawlers and people different content to influence rankings.
• Keyword stuffing — packing a page with keywords or numbers to manipulate rankings.
• Link spam — creating links to or from a site primarily to manipulate rankings.
• Scaled content abuse — many pages made for rankings rather than for users: AI-generated without value for readers, stitched together from other people’s content or from search results.
• Site reputation abuse — third-party content on a well-known site, published to ride on its rankings, such as a third-party author’s payday loan reviews on an educational site. Outside the EEA this can lead to a manual action; since late August 2026, such pages in the EEA are assessed separately from the main site.
• Expired domain abuse — buying an old domain to promote low-value content on the strength of its history.
• Hidden text and links, doorways, sneaky redirects — anything that shows search engines one thing and people another.

AI-written content: Google’s position

AI itself isn’t banned: Google says generative models are useful for researching a topic and structuring original content. The violation is mass-generating pages that give readers nothing — that’s scaled content abuse.

Google advises manually fact-checking everything AI writes, including titles and other metadata: models don’t retrieve facts, they predict a likely sequence of words. If readers might wonder how the content was made, it’s worth telling them. Separate files like llms.txt and special markup for AI features aren’t needed for Google Search.

Local SEO: Google Business Profile

For a business with an address or one that serves customers on site, a Google Business Profile matters: it determines how you show up in local results. Google ranks local results on three criteria: relevance (how well the profile matches the search), distance from the user and prominence (how well-known the business is).

Fill out the profile completely — address, hours, category, attributes such as parking or Wi-Fi — verify your business, keep special hours up to date, reply to reviews and add photos. You can’t buy or request a higher local ranking.

How to measure results

The main report is Search Console’s Performance report: “Clicks,” “Impressions,” “CTR” and “Average position,” broken down by queries, pages, countries and devices. By default it shows the last three months, and Search Console keeps 16 months of data in total. Impressions and clicks from AI Overviews and AI Mode are included in the report’s web search data.

Since August 31, 2026, a separate Generative AI performance report has been available to all sites: it counts only impressions of links to your site in AI Overviews and AI Mode, with no clicks, and it appears once there are enough such impressions.

What visitors do after they arrive — how much they read, where they come from, whether they place an order — is shown by Google Analytics 4 (see our Google Analytics 4 guide). After making changes, wait a few weeks and compare periods of the same length.

SEO checklist for your website

Work through it from top to bottom: technical errors do the most damage, and fixing them doesn’t require rewriting any text.

What to check	Where	Target
Key pages are indexed	Page indexing report, URL Inspection	Status “Indexed”
Sitemap	Sitemaps report	No errors, up to 50,000 URLs / 50 MB
robots.txt	example.com/robots.txt	Doesn’t block needed pages
Canonical URL	URL Inspection	Matches your choice
HTTPS	Browser, URLs in the sitemap	All pages, redirect from HTTP
Mobile version	Phone, URL Inspection	Same content as on desktop
LCP	Core Web Vitals, PageSpeed Insights	≤ 2.5 s
INP	Core Web Vitals, PageSpeed Insights	≤ 200 ms
CLS	Core Web Vitals, PageSpeed Insights	≤ 0.1
Title and description	Page source	Unique, describe the page
Images	Page source	img tag, meaningful alt
Structured data	Rich Results Test	No errors
Paid links	Page source	rel=sponsored or nofollow
Manual actions	Manual actions report	No issues detected
Results	Performance report	More clicks on target queries

Common mistakes

• Blocking a page in robots.txt to remove it from Search. It can stay indexed without a description; you need noindex with crawler access left open.
• Writing to a word count and keyword density. Length alone doesn’t affect rankings, and keyword stuffing violates Google’s policies.
• Publishing AI text without checking it. Models get facts wrong, and hundreds of near-identical pages fall under scaled content abuse.
• Buying “permanent” links. A paid link without rel="sponsored" or nofollow is link spam.
• Expecting results from myths. The keywords meta tag, llms.txt and FAQ markup for snippets do nothing in Google, and keywords in the domain do next to nothing.
• Trusting guarantees. Promises of “#1,” a “special relationship with Google” or “priority submit” — Google explicitly warns to beware of such SEOs.

Sources and methodology

Analysis of public sourcesChecked 9 October 2026

1. Google Search Central — SEO Starter Guide developers.google.com
2. Google Search Central — Creating helpful, reliable, people-first content developers.google.com
3. Google Search Central — Spam policies for Google web search developers.google.com
4. web.dev — Web Vitals web.dev
5. Search Console Help — Performance report (Search results) support.google.com

FAQ

How long does SEO take to work?

Google says some changes show up in Search within a few hours, while others take several months. It suggests waiting a few weeks before judging the effect of your changes.

Can I do SEO for my website myself?

Yes. The core tools are free — Search Console, PageSpeed Insights and the Rich Results Test — and Google publishes an open SEO guide for site owners. If you do hire someone, remember that no one can guarantee the top spot.

Does Google penalize AI-written content?

Not for using AI as such. Google treats mass-generating pages that don’t help readers as a violation, and it advises fact-checking all AI-generated text before publishing.

Do I need special optimization for AI Overviews and AI Mode?

No. According to Google, there are no additional requirements: the page must be indexed and eligible to appear with a snippet. Google Search doesn’t need llms.txt files or special markup.

Does content length affect rankings?

No. Google says content length alone doesn’t matter for ranking, and it calls writing to a particular word count a sign of search engine-first content.