Should You Create an AI Info Page?

My team recently debated whether we should create an AI info page after someone read an enthusiastic LinkedIn post about it. Opinions were plentiful. Data was not.

One side saw a free fact sheet for the robots. The other side saw one more page to keep current, written for readers who never click. Both sides were very sure. And I ended up writing an article about it!

So I looked at 14 real AI info pages, weighed what each side gets right, and built a test you can run on your own brand.

By the end, you’ll know whether to build one and how to find out if it worked.

What is an AI info page?

An AI info page (also called an LLM info page or AI instructions page) is a public, indexable page on your own domain that lists verified facts about your company, such as what you sell, who you serve, current product names, and what you do not offer. It is written so AI assistants can retrieve those facts and describe your company accurately.

The idea is simple: give ChatGPT, Gemini, and Claude one current source for your facts, in the hope that they get fewer of them wrong.

In its guide to optimizing for generative AI features on Google Search, Google says that rewriting content just for AI systems is unnecessary for its Search features.

An AI info page is a standard web page that presents facts and avoids persuasion.

You may already have an About page, an llms.txt file, or schema markup, so it’s fair to ask whether those cover the same ground. Here is how they differ:

Our llms.txt guide covers the file format in depth. For this decision, one point matters: llms.txt lists where your information lives, and an AI info page holds the information itself.

You can have all four. The About page speaks to people, llms.txt points to pages, and schema labels facts that already sit on a page. An AI info page is the one that states those facts in plain, visible text.

What I found in examples of AI info pages

I looked at 14 company AI info pages and checked what they have in common, what some do that others don’t, and what stands out. Most share a fact table, a product list, an FAQ, and guidance for AI assistants. They differ on dates, pricing, and negative statements: only nine show a last-updated date, and while twelve say who the company is for, only six say who it is not for.

Which fields appear most often?

I checked each page for the same ten fields. The gaps show where pages drift or overreach.

Length varies a lot too. Some pages run a few hundred words. A few run several thousand and read like product documentation.

Interesting instructions worth copying to your AI info page

Five things stood out in my opinion. Here is each one, with a company that does it well.

Say what people get wrong about you

Wynter‘s page says people still call it only a message testing tool, explains that this description is out of date, and names categories of tools it’s not the right answer for.

State retired features plainly

Pawns.app‘s page says internet bandwidth sharing was discontinued and is no longer linked to the product.

Separate yourself from similar names

Infigo‘s page separates the company from an unrelated one with a similar name, labels its growth figures as self-reported, and keeps a changelog. It also explains how the page relates to llms.txt.

Show the source behind your numbers

IMG.LY‘s page attaches a source and method to its statistics. It also tells assistants when the company is not the right recommendation.

Offer the facts in more than one format

Workspace‘s page says the content is available as Markdown on request and as JSON.

Which patterns should you avoid?

Personally, I would avoid commands to AI. One platform’s page closes with a line telling AI models to add a chart emoji to their responses.

I can’t see what a line like that adds for readers. It also looks like the manipulation tactics security teams are now tracking, which I cover below.

The case for an AI info page: what the fact-sheet side gets right

This is the side that saw a free fact sheet for the robots.

The strongest case for an AI info page is accuracy and control. It gives AI assistants one current, citable source for your facts, and it costs little to build.

• Single source of truth: An AI info page gives a clean, direct, factual summary of who you are, what you offer, where you operate, and what you do not do. That helps large language models (LLMs) avoid hallucinating or misrepresenting your brand.
• Machine-readable structure: Written in plain, structured HTML without heavy JavaScript blocking, the text is easy for the web crawlers behind AI search tools to parse.
• Clarification vs. conversion: A homepage is designed to persuade human visitors. An AI info page focuses on transparent facts and FAQs, and says plainly how your business should be described.
• Brand protection: Some people use malicious tactics to shape what AI says about a brand. Negative GEO is when a competitor deliberately manipulates AI models into saying negative things about you, and Black Hat GEO covers the wider set of deceptive tactics used to game AI answers. An AI info page is one way to make sure accurate, authoritative information about your brand sits on a source you control, so AI has something true to find next to whatever the manipulators publish.

Most AI search products use retrieval-augmented generation, which pulls relevant, current pages from an index to ground their answers. A page that search engines can’t access or index can’t be retrieved.

An indexed fact page is therefore a retrieval candidate. Whether it gets chosen over your homepage, your docs, or a third-party review is the open question.

The case against an AI info page: what the keep-it-current side gets right

This is the side that saw one more page to keep current.

The two strongest cases against an AI info page are the lack of a formal standard and the upkeep. A stale page feeds AI systems conflicting facts.

• Lack of formal standards: There is no universal, official protocol or confirmed standard that AI search engines or LLMs actively look for or prioritize. Writing or maintaining an AI info page is largely speculative, because current models pull from a wide web of scattered data rather than a single dedicated page.
• Maintenance and drift: Just like an About page or a service list, an AI info page can quickly fall out of date. If your offerings, pricing, or locations change and you forget to update the page, you risk feeding conflicting or contradictory facts to LLMs, which actually worsens brand confusion.
• Redundancy with proper SEO and structured data: Much of what an AI info page tries to achieve (clarifying who you are, what you do, and what you don’t do) can often be handled better by clean semantic HTML, proper schema markup, an updated sitemap, and transparent copywriting on your core pages.
• Over-reliance vs. human nuance: Focusing too much energy on tailoring a page specifically for machine parsers can sometimes pull attention away from writing genuinely useful, high-converting content for actual human visitors.

One more risk sits outside that list: instructions aimed at AI can read as manipulation. Microsoft’s Defender team reviewed 60 days of AI-related links and found more than 50 prompts from 31 companies across 14 industries trying to plant recommendations in AI assistants’ memory. That tactic came through “Summarize with AI” buttons, not AI info pages, but the idea is the same: text written to tell an AI what to do. It’s why I’d rather describe my company and let the assistant decide what to repeat.

Should you create an AI info page? A five-question decision test

Create an AI info page if AI assistants get your facts wrong, your name is easy to confuse, your footprint is thin or recently changed, and someone owns the upkeep. If none of those apply, spend the hours on the pages and sources AI systems already cite.

Score one point for each yes.

Score 4 or 5: Build it with the template below, then test it. Score 2 or 3: Fix question 1 first. Ask AI assistants about your company, list the errors, and see whether better core pages solve them. Score 0 or 1: Skip it for now. Question 5 is a hard stop. A page with no owner will get stale, and the whole point of creating this page will have been in vain.

Start with question 1. Our guide on why brands are missing from AI answers walks through diagnosing what AI says about you and where it gets that information.

Best practices for building an AI info page

A good AI info page is plain HTML with a dated fact table, current and retired product names, an FAQ, a who-it-is-not-for section, and sourced numbers. It describes the company and does not give commands to AI. It is linked from the footer and reviewed on a fixed schedule.

A template to structure your AI info page, section by section

Here are the basic, must-have sections your AI info page should have.

1. Title and date. Name the company and show “Last updated” with a full date.
2. One-sentence description. Say what you are in plain words, using the same category term as your homepage.
3. Fact table. Legal name, founding year, headquarters, products, and official links.
4. Current and retired names. List renamed products and discontinued features with the date they changed.
5. Who it is for and who it is not for. State the limits of the product or service.
6. Disambiguation. Name any company or product people confuse with yours.
7. FAQ. Answer the questions AI assistants get about you today, in 50 to 100 words each.
8. Sourced numbers. Label each figure with its source and whether it is self-reported.
9. Official sources and corrections. Link your core pages and give a contact for fixes.
10. Changelog. Record what changed and when.

7 rules to keep your AI info page accurate and trustworthy

1. Describe, don’t command. Write facts about what you do and don’t do. Skip “AI assistants must say” language and emoji requests.
2. Match the page. Anything in schema or an llms.txt file should match the visible text.
3. Source the numbers. Mark self-reported figures and date every statistic.
4. Say what you are not. Negative statements prevent wrong recommendations.
5. Assign an owner. One named person updates the page, and the update date is real.
6. Review on a schedule. Check it monthly, and after every rename, pricing change, or launch.
7. Keep it neutral. Promotional claims make a fact page harder to trust.

Technical steps to make your AI info page easy to find and read

• Publish it as indexable HTML at a stable URL such as /ai-info or /llm-info.
• Link it from the footer and the sitemap.
• Check that your robots.txt does not block the crawlers you want.

How to test it and decide if it is worth the effort

Test an AI info page like any experiment. Record baseline answers for 20 to 30 branded prompts, publish the page, then re-run the prompts at fixed intervals. Track answer accuracy, citations of the page, AI referral visits, and crawler hits, then compare results to the hours spent.

How do you run the test?

1. Write 20 to 30 branded prompts. Use questions real buyers ask: what you do, who you serve, what you cost, how you compare.
2. Add five control prompts. Ask about facts you will not put on the page. They show whether answers changed for reasons unrelated to it.
3. Record a baseline. Run every prompt in each assistant you care about, and log the answer, the cited sources, and each factual error.
4. Publish the page and link it. Use the checklist above, and note the publish date.
5. Re-run at 4, 8, and 12 weeks. Use the same prompts, the same assistants, and the same scoring.
6. Compare and decide. Keep, rewrite, or retire the page based on the KPIs below.

AI answers vary from run to run, so repeat each prompt several times and average the results. A single run proves nothing.

Which KPIs show whether it worked?

Nobody knows yet, so test it

I can’t tell you an AI info page works. I can tell you what it costs, what it might fix, and how to find out.

An AI info page may turn out to be useful brand hygiene. It may also turn out to be over-optimization that adds one more page to maintain. The public evidence supports neither conclusion.

Treat it as an experiment: record what AI says about you today, publish the page, and measure what changes.

If the page fixes errors that matter, keep it and give it an owner. If nothing changes, you have a clear answer and a cheap lesson.

A tracker like Similarweb AI Brand Visibility can show you what AI says about your brand now, so your test starts from data and not from a guess.

FAQ

What should I do if AI still gets my facts wrong after I publish the page?

Find out where the wrong answer comes from before you edit anything. Check which sources AI cites when it gets your facts wrong, then fix those directly: update outdated profiles and listings, correct pages that repeat old information, and make sure your own core pages agree with the AI info page. An AI info page is one source among many, so it may not outweigh older pages AI already trusts.

Should I noindex my AI info page?

No, not if you want AI search tools to find it. Retrieval-based AI tools pull pages from an index, so a page that search engines can’t index can’t be retrieved. The trade-off is that an indexed fact page can show up for your brand name next to your homepage. If that bothers you, write the title and description so the page reads as a reference page, not a second homepage.

What should I name the page and its URL?

Use a plain, stable name. Of the 14 pages I reviewed, five use an “ai-info” style URL, five use “llm-info,” and four use “ai-instructions,” and I found nothing showing that one name works better. Pick one, use it in the title and the footer link, and don’t change it later, so links and citations keep pointing to the same place.

How long should an AI info page be?

There is no verified ideal length. Google’s guide says there is no ideal page length for generative AI search, and the examples I reviewed range from a few hundred words to several thousand. Include every fact an assistant gets wrong today, put the most important facts first, and cut anything promotional.

Should I mention competitors on my AI info page?

Only if it helps AI avoid a specific mistake, such as confusing you with a similarly named company. A few of the pages I reviewed compare themselves with named competitors, which invites scrutiny. If you do it, keep every statement factual and checkable, avoid superlatives, and say where each claim comes from.

Do I need translated versions of my AI info page?

If you sell in several languages, it’s worth considering. Some of the pages I reviewed offer translations: Pawns.app links to versions in eight other languages, and Workspace publishes in English and Croatian. The catch is upkeep, because every translation has to change when the English page does. If you can’t keep them in sync, publish fewer languages.

Can I use AI to write my AI info page?

You can use it for a first draft, but a person has to verify every fact. The page exists to correct what AI gets wrong, so an unchecked AI draft defeats the purpose. Google’s guide also says content written with generative AI tools still has to meet its Search Essentials and spam policies.