A Public API, No Key, CC-BY. This One Is Built to Be Cited.
Written by Submitator. Infrabase.ai is one of the directories we file customers to, so read it as an interested party. The link numbers were measured from their live pages and the API was queried directly; the footnote says exactly what was read.
Infrabase.ai is a hand-maintained directory of AI infrastructure, rated 26 by Ahrefs, holding 661 URLs: 323 individual product pages across nine categories, 279 alternatives pages, plus comparison pages, a landscape map and a blog. Submitting is free and takes four fields.
The link result is the cleanest kind. We read 80 tool pages and all 80 carry a Visit website anchor straight to the product's own domain with rel="noopener". Excluding their own badge row as furniture leaves 257 external anchors across those pages and every one is followed. No nofollow, no sponsored, no ugc, no redirector, nothing noindexed, and identical under Googlebot.
It also gives you more than one. A product page links your site, your pricing page, your X account and your LinkedIn, all followed. Most directories in this catalog give you a single anchor and call it a listing.
And then there is the part nobody else in this catalog has built at all.
Their robots.txt carries a comment pointing at llms.txt and agents.md. The second of those tells agents, in writing, to "Query the API rather than scraping the HTML pages", and the API is GET /api/query with no authentication, no key, no signup. Every response carries an origin block naming a CC-BY-4.0 licence and an attribution requirement. Your listing is not page copy here. It is a structured record under an open licence that anything can read and is asked to credit.
| Field | Value |
|---|---|
| name | Azure AI Search |
| tagline | Microsoft managed search service with vector, hybrid, and agentic retrieval for RAG |
| categories | vector-databases |
| primary_job | ai-search |
| secondary_jobs | vector-database |
| pricing | Pay-as-you-go |
| hq_country | US |
| gdpr | null |
| github_stars | null |
| site_url | your own canonical URL |
Infrabase.ai at a glance
| What you get | Infrabase.ai |
|---|---|
| Price to submit | $0, four fields, no account |
| The link | Followed and direct on 80 of 80 |
| How many links | Usually four: site, pricing, X, LinkedIn |
| Redirector in the path | None |
| Domain Rating | 26 (Ahrefs) |
| A page of your own | Yes, plus the alternatives pages |
| Machine-readable | llms.txt, agents.md, keyless JSON API |
| Data licence | CC-BY-4.0, attribution required |
| Featured listing | $29/month or $290/year, early-bird |
| Filed for you by Submitator | Free listing, plus 29 more directories from $29 |
Per infrabase.ai, its /submit and /advertise pages, robots.txt, sitemap.xml, llms.txt, agents.md and the /api/query endpoint, all read on 17 September 2026. Domain Rating is Ahrefs' figure via submitator.com/domain-rating-checker, which returned 26. Their robots.txt is "User-agent: * / Allow: /" with a sitemap line and comments pointing at llms.txt and agents.md; nothing is disallowed. The sitemap holds 661 URLs: 323 individual product pages across media-generation, inference-apis, observability-analytics, agents, frameworks-stacks, vector-databases, audio, fine-tuning and prompt-engineering, 279 pages under /alternatives/, 16 blog posts, 9 under /compare/, 8 under /vs/, plus /landscape, /european, /advertise, /about and /submit. The link finding is a sample of pages and a census of anchors within them: 80 tool pages, made up of the first 8 and last 8 in sitemap order plus a seeded random 64, each fetched successfully, with anchors located by regex on the anchor element rather than by an HTML parser. All 80 carry an anchor whose visible text begins "Visit website" pointing at the product's own domain with rel="noopener". Nine external hosts appear on at least 80% of pages and were identified as furniture by frequency before attribution, being Infrabase's own badges on other directories plus github.com; the 257 non-furniture external anchors that remain carry rel="noopener" on 210 and no rel on 47, so none is nofollow, sponsored or ugc, and none of the 80 pages is noindexed. Three pages were then re-requested with a Googlebot user agent and returned the same anchors with the same rel. The four-link count describes the pattern seen on those pages, where a listing links the product site, its pricing page, its X account and its LinkedIn company page; it is not guaranteed and a listing without an X account will carry fewer. The API was queried directly at GET /api/query?category=vector-databases with no key and returned HTTP 200 with an origin block reading source Infrabase.ai, licence CC-BY-4.0 and "Attribution required"; the record fields quoted are those returned for Azure AI Search, and site_url is described generically rather than reproduced. Form fields, the checkbox wording and the line "It's free, no strings attached" are quoted from the live submission form; both checkboxes are optional and the second states "Ticking this commits you to nothing." Featured pricing, placement and the traffic figures, being 1,600 click-throughs in 30 days and over a hundred visitors arriving from ChatGPT, Gemini, Claude or Perplexity, are their own claims from /advertise, quoted rather than verified. Nothing was submitted and no featured listing was purchased, so their review time, acceptance rate and the quality of an extended listing are outside what we can report. Submitator files customers to Infrabase.ai. Our plans are $29 for 30 directories, $49 for 60 and $79 for 100, one-time, chosen by you with each Domain Rating shown and no-equity marked where it applies, badges installed and kept live, rejections credited back as +1 directory and a 30-day refund policy on the conditions in our terms.
How to submit to Infrabase.ai, step by step
- 1
Write the description for the API, not for the page
The field is optional and asks for a sentence or two. Do not skip it. What you write becomes the tagline on a structured record served as JSON under an open licence, which means it is the sentence an assistant is most likely to repeat when somebody asks it to compare tools in your category. Say what the product does and for whom, in one sentence, with no adjectives you would not defend.
- 2
Send the canonical URL, because it becomes site_url
The API returns a site_url field per record, and that is the address anything querying the directory will hand a reader. Give the plain canonical homepage with no campaign parameters, and make sure it is the version you want indexed rather than a redirect that resolves somewhere else.
- 3
Have your X and LinkedIn in order before you file
Listings here carry followed links to the product's pricing page, X account and LinkedIn company page as well as the site. That is three more than most directories give you, and they come from whatever the reviewer can find. An abandoned X account or a missing company page is a link you simply do not get.
- 4
Decide about the link-back box on your own terms
The first checkbox reads "I'd be open to linking back to Infrabase from our site." It is optional, the page above it says free with no strings attached, and the submission is accepted either way. A single reciprocal link between two genuinely related sites is not what Google's policies describe, so this is a goodwill question rather than a risk one, and either answer is defensible.
- 5
Tick the extended-listing box, because it costs nothing
The second checkbox asks for an in-depth overview and FAQ on your product page, and their own note is explicit: "Written by us from your own docs. Most listings never get one... Ticking this commits you to nothing." It is a free expression of interest in a page section somebody else writes for you. There is no reason to leave it blank.
- 6
Make your docs worth writing from
Follow the logic of that last box. If they write the extended listing from your own documentation, then the quality of your docs is the quality of the listing. Whatever you were going to do to your getting-started page this quarter, do it before you submit rather than after.
Why the machine-readable layer matters now
Every directory in this catalog is trying to be found by a search engine. This is the first one we have measured that is visibly trying to be found by an assistant, and it has built the plumbing rather than just saying so: llms.txt as a site map for models, agents.md as a query contract, a keyless endpoint, and a licence that asks to be credited by name.
Their own claim on the advertise page is that over the last thirty days the directory sent 1,600 click-throughs to listed products, and that over a hundred of its visitors arrived from ChatGPT, Gemini, Claude or Perplexity. Those are their numbers and we did not verify them. The infrastructure behind the claim, though, is checkable in about a minute, and it is real.
The practical consequence for a submitter is small but specific. On most directories the thing you type is page copy. Here it is a field in a record with a licence attached, so the tagline you write is the sentence that gets repeated, and the URL you give is the one that gets handed over. Spend the extra two minutes.
Four followed links instead of one
Worth dwelling on because it changes the arithmetic of a DR 26 listing. A product page here links out to the site, the pricing page, the X account and the LinkedIn company page, and all four carry rel="noopener" rather than anything that withholds. Across our sample that pattern held on page after page, with 257 non-furniture external anchors and not one of them marked.
Compare that against directories we have measured at DR 69 and above whose single listing anchor routes through a path their own robots.txt blocks, so nothing crawlable reaches the customer at all. The rating on the card is not the measurement, which is why our catalog prints a Domain Rating and a no-equity mark side by side rather than a rating alone.
What $29 a month buys, and what ours buys
Their featured listing is $29 a month or $290 a year at an early-bird rate. It floats the product to the top of its category with a badge, puts it in the featured block on every /alternatives page in that category, and includes the extended listing. They also offer to send the exact pages and their current click numbers for your category before you decide, which is more disclosure than this market usually offers and which we would take them up on.
Since the number is the same as ours it is worth stating plainly rather than leaving you to notice: their $29 is a month of placement on one directory, renewing. Our $29 is thirty directories, once. They are different purchases and both can be right, but they should be compared on purpose rather than by coincidence.
Infrabase.ai review: the short answer
Free, four fields, a page of your own, four followed links instead of one, and the only directory here with a keyless public API and an open data licence. Write the description as though a model will read it out loud, because one probably will.
Their $29 is one month. Our $29 is thirty directories, once.
One payment, every directory named in your dashboard with its Domain Rating on the card, a no-equity mark where the link cannot be crawled, badges installed and kept live where one is wanted, and a rejection credited back as another directory of your choice.
Four followed links, and a public API.
A DR 26 AI infrastructure directory with 323 product pages and 279 alternatives pages. All 80 listings we read pass a followed link, and a listing usually carries four: site, pricing, X and LinkedIn. It is also the only directory here with llms.txt, agents.md and a keyless JSON API serving CC-BY data, so what you submit becomes a record a model can quote. Featured is $29 a month; ours is thirty directories once. The whole catalog.
FAQ
Is an Infrabase.ai link dofollow?
Yes, on everything we measured. We read 80 of their 323 tool pages and all 80 carry a Visit website anchor pointing straight at the product's own domain with rel="noopener", which is followed. Once their own badge row is excluded as furniture, 257 external anchors remain across those pages and every one of them is followed too: no nofollow, no sponsored, no ugc, no redirector, and nothing noindexed. Three pages re-requested as Googlebot returned the same links.
How many links does an Infrabase.ai listing give you?
Usually four, which is unusual. A product page links to your site, to your pricing page, to your X account and to your LinkedIn company page, and all four are followed. Most directories give you one. That is worth knowing when you compare a DR 26 listing here against a single link somewhere higher rated.
Is Infrabase.ai free to submit to?
Yes. Their own line above the form is “Get listed in the AI infrastructure directory. It's free, no strings attached.” The form asks for a product name, a URL, an optional one-or-two-sentence description and a contact email. There is no account, no badge and no payment step.
Does Infrabase.ai have a public API?
Yes, and it is the most interesting thing about the site. GET https://infrabase.ai/api/query works with no authentication, no key and no signup, filters by category, job, GDPR status, HQ country and text, and returns structured records. Every response carries an origin block naming the CC-BY-4.0 licence and the attribution requirement. Their robots.txt points at it in a comment, and agents.md tells agents to query it rather than scrape the pages.
What data does Infrabase.ai hold about a listing?
Per record: name, slug, tagline, categories, primary and secondary jobs, pricing model, HQ country, GDPR status, GitHub stars, the Infrabase page URL, your own site_url and a last-updated timestamp. That is a lot more structure than a tagline and a logo, and because it is served as JSON under an open licence it is the version of your product an assistant is most likely to repeat.
What is Infrabase.ai's Domain Rating?
26, per Ahrefs. The rating is the least interesting number on this page. The directory holds 661 URLs, of which 323 are individual product pages across nine categories and 279 are alternatives pages, and their advertise page says those alternatives pages carry most of the site's search traffic.
What are the two checkboxes on the Infrabase submission form?
Both optional. The first reads “I'd be open to linking back to Infrabase from our site”, which is an invitation rather than a condition; the page above it says the listing is free with no strings attached. The second reads “I'd like an extended listing: an in-depth overview and FAQ on our page”, with the note “Written by us from your own docs. Most listings never get one... Ticking this commits you to nothing.”
Should I tick the link-back box on Infrabase?
It is your call and it costs nothing to leave blank. A single reciprocal link between two genuinely related sites is not what Google's spam policies describe, which name excessive link exchanges and partner pages existing only for cross-linking. What it does buy is goodwill with a hand-maintained directory, and the submission is accepted either way.
How much is a featured listing on Infrabase.ai?
$29 a month or $290 a year at what they call an early-bird rate. It floats your product to the top of its category with a Featured badge, puts you in the featured block on every alternatives page in that category, and includes the extended listing. They also offer to send the exact pages and their current click numbers for your category before you buy, which is more disclosure than this market usually manages.
Does Submitator submit to Infrabase.ai?
Yes, it is in our catalog with its Domain Rating on the card and the link measured rather than assumed. It is one of the entries where filling the description field properly pays off, because what you write becomes a structured record in a public API rather than a line of page copy, and a rejection comes back as a credit for another directory.
More submit-to guides.
Their $29 is one month. Our $29 is thirty directories, once.
One payment, each directory named with its Domain Rating on the card before you choose it, a no-equity mark where the link cannot be crawled, badges installed and kept live where one is wanted, and a rejection swapped for another directory.