A site search engines find
We check that crawling is not blocked and that the sitemap, titles, descriptions and main addresses are in place.
Free DII check
Enter an address and we check up to 62 things that decide whether users can find your site, in search such as Google and Bing and in AI such as ChatGPT, and score it out of 100. See the score right away; sign up to see each item.
Create a free account to see all 62 results and what to fix first. Add the site and you also get how to fix each item and an llms.txt draft.
Sign up free and see the resultsSign up and register your site, and the console shows how to fix each item and gives you an llms.txt draft written from reading your site. It is in the Free plan.
Start freeAn AI crawler has to open the front page before it can read anything else.
A site without a secure connection is pushed down in search and in AI answers.
If the http address does not move to https, the same site is counted as two.
If the address opens both with and without www, the same site is counted as two.
With noindex, search engines and AI search leave the page out.
The file that tells crawlers what they may read.
A sitemap is how new pages are found quickly.
The sentences search results and AI answers use to introduce the page.
It names the main address when the same content opens at several.
The title and image shown when a link is shared.
The site mark shown in search results and browser tabs.
Registering with Google Search Console and Naver Search Advisor shows how the site is being crawled.
It says which language the page is in.
The fastest way to announce new content.
Crawlers find the other pages by following the links in the HTML.
Search engines judge a site by how it looks on a phone.
Search and AI cannot tell what an image without alt shows.
Images without a stated size push the page around as they load, which lowers the page quality rating.
Scripts that run straight away in the head delay the page appearing.
When the HTML is heavy, crawlers stop before the end and the page appears late.
Crawlers read less of a slow site, and users do not wait.
An uncompressed page takes several times longer to load.
Security headers show a site is looked after, and they protect users from interception.
The format that lets AI read who runs the site and what it is without guessing.
The form that sets the site name shown in search results.
Stating the page topic in a single h1 makes summaries accurate.
Content split under subheadings lets AI pick and quote the part that fits a question.
Skipping a level, such as h3 right after h1, makes the structure of the page read wrongly.
When a question is the heading and the answer follows, that answer is what search and AI show.
The main and article tags are how the content is told apart from the menus.
An AI service that robots.txt blocks cannot read the site when it writes an answer.
If training crawlers are blocked, the next AI models will not know the site. Blocking them is also a choice.
A guide file that tells AI what the site is and what is where.
Many AI crawlers run no scripts, so content drawn only by scripts is not read.
Linking official accounts such as social profiles is how AI recognises them as one brand.
AI leans heavily on Wikipedia. A site Wikipedia cites is treated as reliable.
Search and AI trust a site more when it says who runs it.
A site that shows an email or a phone number reads as one somebody answers for.
Search and AI trust a site more when it publishes its policy and terms.
It is how users find what they want inside the site.
We look for an AI assistant that answers questions from the site's content. A search box alone does not get users all the way to what they want.
It is how one user brings another.
Users come back when there is a place to take part.
It is the most direct way to call back users who have left.
It is a channel for meeting users again with their consent.
Conversion happens when users can see what to do next: sign up, subscribe or buy.
We look for a place that leads to revenue: paid products, membership or ads.
Without measuring, you cannot tell what improved.
Search and AI answers take the gist from the first paragraph.
Sentences with figures and dates are the ones that get quoted.
Text split under subheadings lets the needed part be picked out.
Lists and tables are carried into search results and AI answers as they are.
A subheading written the way users ask matches an AI question directly.
Search and AI trust writing more when it says who wrote it.
A piece is treated as current only when it says when it was written.
The form in which the title, author and date of a piece are read without mistakes.
Structured data that says which section of the site this piece sits in.
Each piece needs its own description and its own address to be listed properly in search.
Linking from the text to related pieces shows the site covers the subject in depth.
Writing that says where its facts come from is what gets quoted.
A piece that is too short reads as not covering its subject, and is not quoted.
A passage is quoted when it makes sense read alone and carries a fact.
Whether search engines find a site, whether it answers questions and whether AI cites it all overlap.
We check that crawling is not blocked and that the sitemap, titles, descriptions and main addresses are in place.
We check what lets content be read as questions and answers, such as structured data and heading structure.
We check that AI crawlers can get in, that there is an llms.txt, and that the content reads without scripts.
publish.link checks your SEO, AEO and GEO and tells you what to improve.
Up to 62 checks, what to improve for each item, and an llms.txt draft written from reading your site.
We draft Organization and WebSite structured data, and score your 5 most recent pieces of content on how ready they are to be quoted in AI answers.
We ask an AI model what it knows about your site, and you print it as a report with the scores of 20 pieces of content.
The check is what we found by opening the front page, robots.txt, sitemap.xml and llms.txt. It does not guarantee that a site appears in AI answers.
GEO is preparing a site so that AI search such as ChatGPT, Perplexity and Google AI Overviews understands it and quotes it in answers. We look at it together with SEO, which is about ranking in search results, and AEO, which is about being the direct answer to a question.
Yes. Enter an address and you see the score and how many things to fix without signing up; the result of each of up to 62 items is there once you sign up free. If you run many checks in a short time, you may have to wait about ten minutes.
We open the front page, one recent piece of content, robots.txt, sitemap.xml and llms.txt of the site, look at up to 62 items, and work out a score out of 100 with a different weight for each item. Only items that pass earn points.
DII is an index that looks at a site's digital competitiveness in ten areas and scores it out of 100. Only part of it can be measured from the address alone, so we show how much was measured as the coverage. Areas that could not be measured are left out of the score, not counted as zero. The same standard applies whichever company's tools a site uses.
Sign up and register your site, and we tell you what to improve for each item. We write drafts of llms.txt and the structured data, so you only need to edit them and put them on your site.
No. Many media block them because they do not want their content used for training. But if you also block the crawlers that fetch pages for answers, your site is unlikely to appear in that AI service's answers. You can decide on training crawlers and answer crawlers separately.
It is not a guarantee. The check looks at the conditions for AI to read your site. Whether you are quoted in an answer also depends on what your content says, so the console scores each piece of content separately.
Check again after you change the structure of the site or robots.txt. In the console you can run the check again at any time.