Let's be honest from the start: nobody controls what AI assistants recommend. There's no form to submit, no paid placement in ordinary answers, and no agency that can guarantee you'll be named. OpenAI says it plainly about ChatGPT search: "Placement is not guaranteed" (OpenAI Help Center). Google says that even pages that meet every requirement aren't guaranteed to be crawled, indexed or shown (Google Search Central).

What you can do is remove the things that stop assistants finding you, and make the facts about your business clear and consistent everywhere they look. That's the part you control, and most small businesses haven't done it yet. Here's a plan in five steps.

Step 1: Find out where you stand

Before changing anything, check what assistants say today. Write six to eight questions a real customer would type: "best [service] in [town]", "[your type of business] near [area] for [specific need]", "alternatives to [a competitor]". Run the same wording in ChatGPT, Gemini and Perplexity, in a fresh chat each time.

For each answer, note whether you were named, who was named instead, whether anything about you was wrong, and which sources were cited. The sources matter most. They show you which websites the assistants trust for your kind of business. Our guide to checking if ChatGPT mentions your brand walks through this in detail.

Save the results with the date. You'll want something to compare against later.

Step 2: Make sure assistants can read your site

This is the most common fixable problem, and it's quick to check.

AI companies run separate bots for search and for training. If you block the search bot, that assistant can't read your pages when it looks things up. OpenAI says sites that opt out of its search crawler, OAI-SearchBot, won't be shown in ChatGPT search answers except as navigational links (OpenAI crawler docs). Perplexity recommends allowing PerplexityBot to appear in its results (Perplexity docs).

You can still opt out of AI training by blocking training bots such as GPTBot. That's a separate setting.

Open yourdomain.com/robots.txt and look for any rule that blocks these search bots, or a blanket Disallow: /. Then check your website host or CDN: some have a "block AI bots" switch that blocks requests before robots.txt is even read. Our robots.txt guide has copy-paste settings.

For Google's AI Overviews and AI Mode, Google says the normal rules apply: your page needs to be indexed and eligible to appear in Google Search with a snippet, and there are no extra requirements (Google Search Central).

Step 3: Say plainly what you do, where and for whom

Assistants are trying to match a question to a business. Make that match easy.

Put one plain, factual paragraph near the top of your homepage or About page:

[Business name] is a [what you are] based in [town], serving [areas or customer type] since [year]. We [top two or three services]. [One true, checkable fact, such as a trade registration.] Contact: [phone], [email], [address or "online only"].

No slogans. "Award-winning solutions" tells an assistant nothing; "Gas Safe registered boiler repairs in north Bristol" tells it a lot.

Then build a page for each question customers ask you most, such as "How much does a boiler service cost?", "Do you cover [area]?" or "How long does [job] take?". Answer in the first two sentences, then add detail. A page shaped like an answer is easier to quote than a page shaped like a brochure.

On structured data: it can help machines read your facts, but Google says you don't need special schema, AI text files or markup to appear in its AI features (Google Search Central). Treat it as tidying up, not a trick. The same goes for llms.txt, which is a voluntary proposal; do the basics first.

Step 4: Get your facts consistent off your site

Much of what an assistant knows about a small business comes from other sites: maps listings, directories, review platforms, trade registers and "best of" articles. If they disagree with each other, for example an old phone number on one, a different business name on another, you look less certain.

Go through them one at a time:

• Google Business Profile. Google's AI features guide specifically recommends keeping your Business Profile information up to date (Google Search Central).
• Bing Places for Business. ChatGPT search uses third-party search providers, and OpenAI's help page points to Microsoft's privacy statement among them (OpenAI Help Center). Bing Places is free to claim.
• Apple Business, Yelp, Facebook, and any review site your customers use.
• Your trade body or professional register, such as Gas Safe, NICEIC, the FCA register or the Law Society. These are high-trust sources.
• Local directories: your chamber of commerce or council business listings.

Make your name, address, phone number and opening hours identical everywhere.

Then search for the questions your customers ask, such as "best [service] in [town]", and open the top few results. If a "best of" page keeps coming up, and it accepts suggestions, ask to be considered. Being mentioned on a page assistants already cite can matter more than another homepage tweak.

Step 5: Ask for honest reviews, the right way

Reviews that describe the job and the place help people and assistants alike. Ask every customer, not only the happy ones, with a short message and a direct link. Don't offer discounts or gifts in exchange, don't write reviews yourself and don't script what customers say. In the UK, fake reviews and reviews that hide an incentive are banned under the Digital Markets, Competition and Consumers Act 2024 (CMA guidance).

Then re-check monthly

AI answers vary from run to run, so one check tells you little. Once a month, run the same questions in the same assistants and record the results in the same format. Look for trends over two or three months, and note what you changed in between. If you track visits, OpenAI says ChatGPT search adds utm_source=chatgpt.com to referral links (OpenAI Publishers FAQ), so you can see whether anyone arrives that way.

What to avoid

• Anyone selling "guaranteed ChatGPT rankings" or "AI placements".
• Fake reviews, fake forum posts or networks of accounts talking you up.
• Stuffing pages with keywords or hidden text aimed at bots.
• Judging success or failure on one answer.

If you want the whole kit

You can do all of this for free with the steps above. If you'd rather work from ready-made files, the Crux AI Visibility Bundle (£15) puts our three self-serve tools together: the Starter Checklist for your first snapshot (step 1), the Fix Kit with robots.txt, llms.txt, schema and listings templates (steps 2 to 5), and the Tracker spreadsheet for the monthly re-check: https://cruxstudio5.gumroad.com/l/bundle. They help you check, fix the basics and measure. They can't make an assistant recommend you, and nor can anyone else. Questions: cruxstudio@mail.grokbot.com

You don't need a monitoring tool to find out. You need a list of good questions, three free AI assistants and about an hour. Here is the method we use, step by step.

Before you start: what "mentioned" really means

AI answers are not a ranking. The same question can produce a different list on a different day, with different wording, or for a different person. So you are not looking for a position number. You are looking for a pattern: across a set of realistic questions, how often are you named, how accurately are you described, and who gets named instead?

Step 1: Write down the basics (10 minutes)

Before you type a single prompt, note:

• your brand name, exactly as buyers would spell it
• your category in plain English (e.g. "online booking software for hair salons", not "next-gen scheduling platform")
• your ideal buyer and what they are trying to get done
• two or three competitors you actually lose deals to
• your market (UK, a city, a niche such as agencies or ecommerce)
• today's date

Your prompts should sound like your buyer, not your marketing.

Step 2: Write six to eight buyer-style prompts (10 minutes)

The biggest mistake is asking "What do you know about [my brand]?". That tells you whether the model has heard of you, not whether it recommends you. Buyers rarely ask that way. Write questions they would type instead, across a few types:

• Category: "Best [category] for [use case]?" or "How should a [role] choose a [category] tool?"
• Comparison: "[Competitor A] vs [Competitor B]: which is better for [use case]?"
• Alternatives: "What are the best alternatives to [Competitor A]?"
• Brand-aware: "Is [your brand] a good option for [use case]?"
• Objections: "Pros and cons of [your brand]."
• Local or segment: "Best [service] in [town] for [specific need]."

Pick six to eight. Mostly category, comparison and alternatives questions, because those are the ones where you either make the shortlist or you don't. Write the exact wording down: you will reuse it word for word.

Step 3: Run the same prompts in three assistants (25 minutes)

Use ChatGPT, Google Gemini and Perplexity. They don't all draw on the same sources, so one tool on its own gives you a partial picture.

A few rules keep the results fair:

1. Start a fresh chat for each prompt, or at least a new conversation per tool, so earlier questions don't steer later answers.
2. Paste the exact same wording in every tool. Don't rephrase.
3. Don't coach the model towards your brand. No follow-ups like "what about [us]?" until you've recorded the first answer.
4. Note the model or mode each tool shows, and whether web search was used. Perplexity usually shows its sources; ChatGPT shows them when it searches the web.
5. If you have time, run each prompt two or three times. If you're named in two runs out of three, that tells you more than one lucky (or unlucky) answer.

Step 4: Record four things per answer (included in the 25 minutes)

Make a simple table, one row per prompt, with a column per assistant. For each answer, note:

1. Were you named? Yes or no, and roughly where in the answer.
2. Who else was named? Write down every brand.
3. Is anything about you wrong? Old pricing, a feature you don't have, the wrong location, a confusion with another company.
4. Which sources were cited? Review sites, directories, "best of" articles, Reddit threads, your own site, a competitor's blog.

The sources column is often the most useful: it shows where the assistants get their information.

Step 5: Look for patterns, not verdicts (10 minutes)

Now step back and score it informally:

• On how many prompts were you named, out of the total?
• On how many were only competitors named?
• Which competitor shows up most, and on which type of question?
• Did any answer describe you wrongly?
• Which sources come up again and again?

Some patterns we look for:

• You're named on brand questions but not on category questions. The assistant knows you exist but doesn't connect you with the problem you solve. Your site may not say plainly what you are and who you're for.
• Answers about you are vague or wrong. Your About page, pricing page and FAQ may need firmer, checkable facts, and your listings elsewhere may disagree with each other.
• The same third-party pages keep getting cited. Those review sites, directories and comparison articles are part of your visibility now. Check whether you're listed on them and whether the details are accurate.
• You're not named anywhere, even on brand questions. Before you worry about content, check that AI search crawlers can actually read your site (see our robots.txt guide).

Step 6: Pick three things to do this week (5 minutes)

Don't try to fix everything. Pick three concrete actions based on what you saw, for example:

1. Rewrite your homepage's first paragraph so it says, in plain words, what you do, for whom and where.
2. Correct one listing or review profile that an assistant cited with wrong details.
3. Write one page that directly answers a question where only competitors were named.

Then save your table with the date. Run the same prompts again in a month and compare like with like.

What not to do

• Don't buy "AI placements" or "guaranteed ChatGPT rankings". Nobody controls what AI assistants recommend, including us. Anyone promising otherwise is guessing.
• Don't post fake reviews or ask friends to flood forums with mentions. It breaks platform rules, and in the UK fake reviews are banned under the Digital Markets, Competition and Consumers Act 2024 (CMA guidance).
• Don't spam assistants with prompts to "train" them. It doesn't work that way.
• Don't treat one answer as proof that something worked, or failed.

A note on tracking traffic from ChatGPT

If people click through from ChatGPT search, OpenAI says it adds utm_source=chatgpt.com to referral links, so you can see those visits in Google Analytics or similar (OpenAI Publishers FAQ).

If you want the checklist version

Everything above is enough to run your first check today. If you'd rather not build the prompt list and table from scratch, the Crux AI Visibility Starter Checklist puts this method in one short file (about 2–4 printed pages, PDF and Markdown): a fill-in prompt bank, a ready-made results table for ChatGPT, Gemini and Perplexity, a 60-minute run sheet and a guide to sensible next steps. It's pay-what-you-want on Gumroad: https://cruxstudio5.gumroad.com/l/ftesfj. It's a starter checklist, not software, and it won't promise you mentions. Questions: cruxstudio@mail.grokbot.com

Most AI companies now run separate bots for separate jobs. Once you know which is which, you can opt out of training and still stay visible in search. Here's the breakdown, checked against each company's own documentation.

The three kinds of AI bot

It helps to sort them into three groups:

1. Training crawlers collect public web content that may be used to train AI models. Blocking them is about whether your content goes into future models. It doesn't directly decide whether you appear in answers.
2. Search crawlers index pages so an assistant can find and cite them when it searches the web. Block these and you may drop out of that assistant's search answers.
3. User-triggered fetchers visit a page because a person asked the assistant something that needs it. They don't crawl the web on their own, and some companies say robots.txt may not apply to them.

OpenAI (ChatGPT)

OpenAI's crawler documentation lists these user agents:

• GPTBot: training. OpenAI says GPTBot crawls content that may be used in training its generative AI foundation models, and that disallowing it indicates your content shouldn't be used for training.
• OAI-SearchBot: search. It's used to surface websites in ChatGPT's search features. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links.
• ChatGPT-User: user-triggered. Used when a user's question in ChatGPT or a Custom GPT leads it to visit a page. OpenAI says it isn't used for automatic crawling, that robots.txt rules may not apply because a user started the action, and that it isn't used to decide whether content appears in search.
• OAI-AdsBot: ads review. It only visits landing pages submitted as ChatGPT ads. Most sites can ignore it.

Two details from the same page are worth knowing. The settings are independent, so you can allow OAI-SearchBot and disallow GPTBot. And OpenAI says it can take about 24 hours after a robots.txt change for search results to adjust.

Example user-agent strings, as published by OpenAI (version numbers may change):

GPTBot:        Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; GPTBot/1.4; +https://openai.com/gptbot
OAI-SearchBot: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot
ChatGPT-User:  Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); compatible; ChatGPT-User/1.0; +https://openai.com/bot

OpenAI also publishes IP ranges for each bot (for example https://openai.com/searchbot.json), which matters if your firewall needs an allow rule.

Perplexity

Perplexity's crawler page lists two:

• PerplexityBot: search. It surfaces and links websites in Perplexity's search results. Perplexity says it is not used to crawl content for AI foundation models.
• Perplexity-User: user-triggered. It visits pages when a user asks a question. Perplexity says it isn't used for web crawling or training, and that because a user requested the fetch, it generally ignores robots.txt.

Perplexity doesn't list a separate training crawler. It says changes can take up to 24 hours to show.

Anthropic (Claude)

Anthropic's help article describes three bots:

• ClaudeBot: training. Restricting it signals that your site's future content should be excluded from Anthropic's model training datasets.
• Claude-SearchBot: search. Blocking it stops your content being indexed for search, which Anthropic says may reduce your visibility in search results.
• Claude-User: user-triggered. Blocking it stops Claude fetching your pages when a user asks, which Anthropic says may reduce your visibility for user-directed web search.

Anthropic says its bots honour robots.txt and that you need to add the rules for each subdomain. It also warns that blocking its IP addresses may not work reliably as an opt-out.

Google (Gemini, AI Overviews)

Google works differently. Google-Extended isn't a separate crawler; it's a robots.txt token. It controls whether content Google crawls can be used to train future Gemini models and for grounding in Gemini Apps and Vertex AI. Google says it does not affect inclusion in Google Search and isn't a ranking signal.

AI Overviews and AI Mode in Google Search are controlled by normal Googlebot rules. Google's AI features guide says there are no extra technical requirements to appear in them, and that robots.txt rules for Googlebot are the control for Search.

So blocking Google-Extended won't remove you from AI Overviews, but it may affect grounding in the Gemini app. Weigh that before you add it.

A starting robots.txt

If you want to stay visible in AI search but opt out of training, this is a sensible base. Merge it with your existing rules; don't paste it over them.

# AI search crawlers: allow, to stay eligible for AI search answers
User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

# AI training: opt out (your choice)
User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

# Google-Extended also affects Gemini app grounding; decide deliberately
# User-agent: Google-Extended
# Disallow: /

User-triggered fetchers (ChatGPT-User, Perplexity-User, Claude-User) are left out on purpose. Only add rules for them if you understand the trade-off.

Check your firewall as well as robots.txt

robots.txt is a request, not a lock. A firewall or CDN can block a bot before it ever reads the file. OpenAI recommends allowing requests from its published IP ranges (OpenAI docs), and Perplexity gives Cloudflare and AWS WAF set-up steps (Perplexity docs). If your host or CDN has a "block AI bots" setting, check what it actually blocks.

A 10-minute audit

1. Open https://yourdomain.com/robots.txt.
2. Search for OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot, and any User-agent: * group with Disallow: /.
3. If a search bot is blocked and you didn't mean it, fix it. Allow about 24 hours for OpenAI and Perplexity to catch up.
4. Check your CDN, firewall or security plugin for AI-bot blocking.
5. On Shopify, Wix or Squarespace, check the platform's own help pages before editing.

What this won't do

Allowing search crawlers doesn't make any assistant recommend you. Nobody controls that. It only makes sure you aren't shut out. What helps you get named is mostly clear facts on your site, consistent business details elsewhere and honest reviews.

If you want the files ready-made

The audit above is complete on its own. If you'd rather start from templates, the Crux AI Visibility Fix Kit (£9) includes a robots.txt starter that separates AI search bots from training bots, plus a 5-page guide with eight fixes in priority order, llms.txt and JSON-LD schema templates, a listings consistency sheet, an answer-first page template, review-request scripts and a monthly re-check routine: https://cruxstudio5.gumroad.com/l/fix-kit. It won't guarantee AI mentions; it fixes the basics you control. Questions: cruxstudio@mail.grokbot.com

Often the cause is a single file. Here's how to check yours in about ten minutes.

Search bots vs training bots

Most AI companies run separate crawlers for answering questions and for training models. You can allow one and block the other.

Sources: OpenAI crawler docs, Perplexity bots, Anthropic crawler help, Google common crawlers.

Three details worth knowing:

• OpenAI says sites that opt out of OAI-SearchBot won't be shown in ChatGPT search answers, though they can still appear as navigational links. It also says robots.txt changes take about 24 hours to apply.
• GPTBot is for training only. Blocking it doesn't remove you from ChatGPT search.
• User-triggered agents such as ChatGPT-User and Perplexity-User act on behalf of a person, so robots.txt may not apply to them in the same way. Don't rely on those entries to control search inclusion. The search bots are the ones that count.

A common way people break this: someone adds User-agent: GPTBot / Disallow: / to opt out of training. Later a plugin or hosting setting adds a blanket "block AI bots" rule that also catches OAI-SearchBot. The site drops out of ChatGPT search and nobody notices.

The 10-minute check

1. Open https://yourdomain.com/robots.txt in a browser.

2. Search the file for OAI-SearchBot, PerplexityBot, Claude-SearchBot, Bingbot, and any User-agent: * block that contains Disallow: /.

3. If a search bot is disallowed and you didn't mean it to be, fix it. A minimal starting point:

   User-agent: OAI-SearchBot
   Allow: /
   
   User-agent: PerplexityBot
   Allow: /
   
   User-agent: Claude-SearchBot
   Allow: /
   
   # Optional: opt out of model training only
   User-agent: GPTBot
   Disallow: /
   
   Sitemap: https://yourdomain.com/sitemap.xml
   

4. Check your CDN or firewall. robots.txt is a request, not enforcement. Cloudflare and some hosts have "block AI bots" or bot-fight toggles that block requests before robots.txt is ever read. If one is on, make sure search crawlers are still let through. OpenAI publishes its search crawler IP ranges (searchbot.json), Perplexity documents WAF allowlisting on its bots page, and OpenAI has guidance on firewall, CDN and captcha layers.

5. Check what the page actually serves. If your main content only appears after JavaScript runs, or the first load is a cookie wall or challenge page, a crawler may get an empty shell even with a 200 status. As a quick sanity check, run curl -s https://yourdomain.com/your-page | less and confirm your key text is in the raw HTML. Changing the user-agent in curl only simulates a bot. Your server logs are the real evidence of what a verified crawler received.

6. On Shopify, Wix or Squarespace, the platform manages robots.txt. Read the platform's docs before you edit it.

What about llms.txt?

It's cheap to add, but don't expect much from it yet. An Ahrefs study of 137K sites found 97% of llms.txt files received zero fetches. An analysis of 300K domains found no clear effect on AI citations. Google says its Search doesn't use it. Treat it as optional, and only after robots.txt, firewall and page content are right.

What this won't do

Unblocking crawlers doesn't make an assistant recommend you. It only puts you back in the pool. What gets you named mostly happens off your site:

• consistent business details across Google Business Profile, Bing Places and directories
• honest reviews
• appearing in the "best of" and comparison pages that assistants already read and cite

Bing is worth a look too. ChatGPT search draws partly on Bing's index, and one study found Bing rankings correlated with ChatGPT visibility. So make sure you're indexed in Bing Webmaster Tools.

No one can guarantee AI mentions or rankings. Anyone who does is guessing.

Quick wins after robots.txt

• Put one plain, factual paragraph on your homepage: who you are, where you are, what you do, since when, and how to contact you.
• Make your name, address and phone number identical everywhere they appear.
• Ask every customer, not just the happy ones, for an honest review, with no incentives.
• Write down 6–8 questions a real buyer would ask. Ask them in ChatGPT, Perplexity and Gemini once a month, and note who gets named and which sources get cited.

Disclosure: we're Crux and we made the Fix Kit. If you'd rather not assemble all this yourself, the AI Visibility Fix Kit (£9) has a robots.txt starter, llms.txt and JSON-LD schema templates, a listings consistency sheet, an answer-first page template, review-request scripts and a monthly re-check routine. For just the buyer-question test from the last bullet, there's a £5 checklist. The check above is everything you need to get started, though. Questions: cruxstudio@mail.grokbot.com