How can I check whether ChatGPT, Gemini and Perplexity recommend my business?

How to Check Whether ChatGPT, Gemini, and Perplexity Recommend Your Business

To check whether ChatGPT, Gemini, and Perplexity recommend your business, ask each one the questions your customers ask, in a fresh session with no account history, and record whether your business is named, what the assistant says, and which sources it cites. Repeat across 15–30 prompts and score the results. WorkspaceCMS is where you publish the fixes once you see the gaps.

That is the whole method. The rest of this guide shows you how to run it properly: which prompts to use, how each assistant behaves, how to score what you find, and what to change on your website when the answers come back empty or wrong. You need a spreadsheet, about 90 minutes, and a willingness to read what the assistants say about you.

The Five-Step Check at a Glance

If you only read one section, read this one. Here is the full process in order.

1. Write 15–30 prompts that mirror how real customers ask for a business like yours. Mix category searches, location searches, comparison questions, and direct brand questions.
2. Run every prompt in all three assistants using a logged-out or fresh session, so personal history does not skew results.
3. Record four things per answer: whether you are mentioned, your position in any list, the accuracy of what is said, and the sources cited.
4. Score and sort the results into mentioned and accurate, mentioned and wrong, and absent.
5. Fix the cause on your own site, then re-run the same prompts in 30 days and compare.

Most businesses skip step three and step five. They glance at one answer, feel good or bad, and move on. A single answer tells you almost nothing because these tools vary their wording from one run to the next. A structured set of prompts, recorded and repeated, tells you where you stand.

Why You Cannot Trust a Single Answer

AI assistants generate answers fresh each time. Ask the same question twice and you can get two different lists. That is normal behavior, not a bug. It means one lucky mention or one unlucky omission is not a result.

Three factors cause the variation:

• Wording changes. "Best accountant in Leeds" and "who should I hire for small business accounting in Leeds" pull different answers.
• Session context. A logged-in account carries memory, location signals, and past conversations. Those shape what you see.
• Retrieval differences. Some assistants search the live web for each question. Others lean more on what they learned in training. The mix changes the sources and the recommendations.

The fix is volume and repetition. Run enough prompts that patterns show up. A business named in 12 of 20 prompts has a real presence. A business named in one of 20 does not, even if that one mention was flattering.

How These Assistants Decide Who to Recommend

Understanding the mechanics tells you what to look for in your results. At a practical level, an assistant recommends a business when it has found enough consistent, clear information to describe it with confidence.

What feeds an assistant's answer

Three kinds of input matter most:

• Your own website. Clear service pages, location details, pricing context, and plain descriptions of who you serve give the assistant material it can quote.
• Third-party mentions. Directories, review sites, news articles, industry lists, and forum threads that name you confirm that you exist and what you do.
• Consistency across both. If your site says you are a "web design studio" and every directory says "marketing agency," the assistant has a muddier picture to work from.

Live search versus learned knowledge

Perplexity is built around live retrieval. It searches, reads pages, and shows numbered citations beside its answer. ChatGPT and Gemini combine learned knowledge with optional or automatic web lookups, depending on the question and the product settings at the time. That is why you test all three. A business can appear in Perplexity because a recent page ranked well, yet be missing from ChatGPT because the assistant leaned on older learned knowledge for that prompt.

You do not need to know the internals. You need to observe the outputs and trace them back to what exists about you online.

Before You Start: Set Up a Clean Test

Bad setup produces misleading results. Spend 10 minutes on this and the rest of the process gets more reliable.

Use fresh sessions

Open a private or incognito browser window. If the assistant works without logging in, use it that way. If it requires login, create a test account you have not used for anything else, or at least start a new conversation and turn off memory features where the product allows it. You want the answer a stranger would get, not the answer tailored to your history of searching for your own company.

Set the location deliberately

For local businesses, location drives everything. Include the city, region, or neighborhood in the prompt text rather than relying on device location. This makes your tests repeatable across days and across the people on your team who run them.

Build a recording sheet

Create a spreadsheet with these columns:

• Prompt text
• Assistant (ChatGPT, Gemini, or Perplexity)
• Date and time
• Mentioned? (yes or no)
• Position (first, second, third, or "not ranked")
• Accuracy (correct, partly wrong, wrong)
• Sources cited
• Notes

That is 8 columns. Resist the urge to simplify. The sources column in particular pays off later, because it shows you which pages the assistants trust.

Build Your Prompt List

The prompt list is the heart of the check. Aim for 15–30 prompts across five categories. Write them the way a customer speaks, not the way a marketer writes.

Category 1: Category and need prompts

These describe the problem or service without naming anyone. They reveal whether you appear for unbranded discovery, which is where new customers come from.

• "What is the best [service] for a small business?"
• "Who offers [service] for [type of customer]?"
• "I need [outcome]. What companies can help?"

Category 2: Location prompts

Add a place. These matter most for local and regional businesses.

• "Best [service] in [city]"
• "Top-rated [service] providers near [neighborhood]"
• "[Service] companies in [region] that work with [customer type]"

Category 3: Comparison and shortlist prompts

These mimic the buyer who is already deciding. Write them in generic terms so the assistant builds a list on its own, then see whether you are on it.

• "What should I look for when choosing a [service] provider?" (then check whether it names you in examples)
• "Give me a shortlist of 5 [service] providers in [city] and explain the differences."
• "Which [service] provider is best for [specific need]?"

Category 4: Direct brand prompts

Now ask about yourself by name. This checks whether the assistant knows you exist and describes you accurately.

• "What is [your business name]?"
• "What does [your business name] do?"
• "Is [your business name] a good choice for [customer type]?"
• "What do customers say about [your business name]?"

Category 5: Problem and question prompts

Customers often ask a question before they know they need a provider. Appearing in these answers is the strongest sign of authority.

• "How do I [task your service solves]?"
• "Why is my [problem] happening and who can fix it?"
• "What does [service] cost and what affects the price?"

A sample split for 20 prompts

• 5 category and need prompts
• 5 location prompts
• 4 comparison prompts
• 3 direct brand prompts
• 3 problem and question prompts

Multiply by three assistants and you have 60 tests. At about 90 seconds each, that is roughly 90 minutes of work. Split it across two people and it takes under an hour.

Checking ChatGPT

Open ChatGPT in a private window or fresh conversation. Paste your first prompt exactly as written in your sheet.

What to look for in ChatGPT

• Named mention. Is your business listed by name? Note the position if it gives a ranked list.
• Description accuracy. Check services, location, and audience. Assistants sometimes blend details from similarly named companies.
• Source links. When ChatGPT searches the web, it can display sources. Record every link. If it does not search, note "no sources."
• Hedged or generic answers. If it says "I don't have specific information about that company," your direct brand presence is thin.

Run a follow-up

After the first answer, ask: "Which sources did you use for that?" or "Can you check the web for more recent options?" The follow-up shows whether a lookup surfaces you when the first pass does not. Record both results separately. A business that appears only after a prompted web search has a retrieval opportunity, since the information exists online but is not strong enough to surface by default.

Checking Gemini

Gemini is Google's assistant, and it often draws on Google's search index and related data. Run each prompt in a fresh session the same way.

What to look for in Gemini

• Local and map-style details. For location prompts, Gemini often reflects business profile data such as hours, address, and ratings. If your Google business listing is incomplete, you will see it here.
• Consistency with search. Compare Gemini's answer against a normal Google search for the same query. Large gaps point to a content or structured-data problem.
• Factual errors. Wrong opening hours, old addresses, or discontinued services are common when source data is stale.

Use the double-check feature carefully

Gemini offers ways to verify claims against web results. Use that on any answer that mentions you. If a statement about your business cannot be verified, trace where the assistant might have picked it up, then correct the source.

Checking Perplexity

Perplexity makes the audit easiest because it shows numbered citations next to each claim. Run each prompt and read the source list before you read the answer.

What to look for in Perplexity

• Citation presence. Is your domain among the cited sources, even if your name is not in the written answer? A citation without a mention is a partial win.
• Which pages get cited. If it cites your homepage but not your service pages, your service pages are unclear or thin.
• Who gets cited instead. Note the directories, publications, and review sites that appear. These are the third-party places you need to be listed on.

Try the focus modes

Perplexity lets you narrow the kinds of sources it searches. Run your top 5 prompts in its default mode, then again with a narrower source setting where available. If you appear in one and not the other, you learn which type of source supports you.

The Three Assistants Compared

Here is how the three differ for the purposes of this audit.

Features and settings in these products change often, so treat the table as a guide to where to look, and check each product's current options when you run your audit.

Score Your Results

Raw notes become useful once you score them. Use a simple scale per answer.

• 3 points: Named, accurate, and positioned in the first three recommendations.
• 2 points: Named and accurate, but lower in the list or described briefly.
• 1 point: Named but with wrong or outdated details, or only cited as a source without a name mention.
• 0 points: Not mentioned at all.

Add up points per assistant and divide by the maximum possible. With 20 prompts, the maximum is 60 per assistant. If ChatGPT scores 21, that is 35%. If Perplexity scores 33, that is 55%. The percentages give you a baseline to beat next month.

Break scores down by category

Also total scores by prompt category. A common result: strong on direct brand prompts, weak on category prompts. That pattern means people who already know you can find you, but strangers cannot. Each category pattern points to a different fix, covered below.

Reading the Results: Five Patterns and What They Mean

Pattern 1: Absent everywhere, even on brand prompts

The assistants do not know you exist. Your site may be new, thin, or poorly indexed, and few outside sources mention you. Start with the basics: a clear homepage stating what you do, where, and for whom; dedicated pages for each service; and listings on the main directories in your field.

Pattern 2: Present on brand prompts, absent on category prompts

The assistants know you but do not connect you to the need. Your pages describe you, but they do not answer the questions customers ask. Publish content that addresses those questions directly, with your services as the practical answer.

Pattern 3: Mentioned with wrong details

This is a consistency problem. Find the source of the wrong information. Perplexity's citations usually reveal it. Fix your own pages first, then update the directory or profile that carries the error.

Pattern 4: Strong in one assistant, weak in others

Each assistant leans on different data. If you win in Perplexity but lose in Gemini, your web content is solid but your business profile data is probably incomplete. If you win in Gemini but lose in ChatGPT, third-party mentions and brand clarity are likely the gap.

Pattern 5: Cited but not named

Your pages inform the answer, but the assistant presents the information without crediting you. Make your brand name more prominent in the content itself. State who you are and what you offer in the opening lines of key pages, so the connection between the information and your business is explicit.

A Worked Example

The following is an illustrative scenario, not a real client result. A small bookkeeping firm in Manchester runs 20 prompts across the three assistants.

• On the 3 direct brand prompts, all three assistants describe the firm correctly. Score: 9 of 9 each.
• On 5 location prompts such as "best bookkeeper in Manchester for freelancers," the firm appears twice in Perplexity, once in Gemini, and never in ChatGPT.
• On 5 category prompts, it never appears.
• Perplexity's citations for the location prompts show two local directories and a regional business blog. The firm's own site is never cited.

The diagnosis is clear. The firm has a brand footprint but no content that matches how people search. Its site has a single "Services" page that lists everything in one paragraph. The fix: separate pages for freelancer bookkeeping, small company bookkeeping, and payroll support; a clear location statement; answers to common cost and process questions; and listings on the two directories that Perplexity trusts. After 30 days, the firm re-runs the same 20 prompts and compares scores.

Fixing the Gaps on Your Own Website

Once you know where you are missing, the work happens on your site. These are the changes with the clearest effect on whether assistants can find, understand, and quote you.

Write a plain description of your business

Your homepage should state in the first two sentences what you do, who for, and where. No slogans. Assistants quote plain statements. They skip vague ones.

Give each service its own page

One page per service, each with a clear title, a direct opening answer, details of who it suits, and common questions. A single catch-all services page gives an assistant too little to cite for any specific need.

Answer real customer questions

Use the problem and question prompts from your audit as a content plan. Each question that returned nothing for you is a page or post to write. Open each with a direct answer, then support it with specifics.

Add FAQ sections

Short question-and-answer blocks mirror how people prompt assistants. Keep each answer to 2–4 sentences and make it self-contained.

Keep facts consistent everywhere

Business name, address, phone number, service list, and hours should match across your site, your business profiles, and every directory. Conflicts reduce confidence and produce wrong answers.

Earn third-party mentions

Assistants cross-check. Get listed in the directories and industry sites that Perplexity cites for your category. Ask for reviews on the platforms your customers use. Contribute to local or trade publications where you have something useful to say.

Where WorkspaceCMS Fits in This Process

WorkspaceCMS does not run the assistant checks for you. You run those in ChatGPT, Gemini, and Perplexity with the method above. WorkspaceCMS is the all-in-one place where you act on what you find: build the pages, write the content, optimize it, and publish it from one dashboard.

Here is how its built-in tools map to the fixes above.

• Service pages and site management. Create and edit pages for each service yourself, without waiting on a developer for a headline change.
• AI-assisted writing. Draft answers to the customer questions your audit exposed, then edit them so they carry your specifics.
• Built-in SEO scoring. Check each page as you write, so titles, descriptions, and structure are in place before you publish. The platform gives you an SEO foundation out of the box, with no plugins needed.
• Visual content calendar. Plan a month of answer-style posts based on your gap list, then schedule and forget. Posts publish automatically.
• Forms on every plan. Capture the enquiries that arrive once assistants start sending people your way. Booking and scheduling are available on the Growth and Premium plans.
• Headless API. If your content needs to feed another site or app, the API is there when you need it.

One clear limit: ongoing SEO and AI campaigns are a separate add-on, not part of a website plan. What the plan gives you is the SEO foundation and the publishing tools. The thinking about what to publish comes from your audit results.

A 30-Day Plan After Your First Audit

Week 1: Fix the facts

• Correct any wrong details found in the audit on your own site.
• Update your business profiles and the directories that carry errors.
• Rewrite your homepage opening to state what you do, for whom, and where.

Week 2: Build service pages

• Create or rewrite one page per service, each opening with a direct answer.
• Add a short FAQ to each page.
• Run each page through SEO scoring before publishing.

Week 3: Publish answer content

• Pick the 4 category or question prompts where you scored lowest.
• Write one post for each, with a direct answer in the first paragraph and specific examples.
• Schedule them across the following two weeks in your content calendar.

Week 4: Earn mentions and re-test

• Submit your business to 3–5 directories or lists the assistants cited.
• Request reviews from recent customers on the platforms that matter in your field.
• Re-run the same prompts. Compare scores against your baseline.

Expect gradual movement. Some assistants pick up new pages within days, while others take longer to reflect changes. Judge progress over two or three monthly cycles, not one.

Common Mistakes When Checking AI Recommendations

• Testing once. One answer is an anecdote. Use at least 15 prompts.
• Testing only your brand name. Brand prompts show whether you are known. Category and location prompts show whether you are found.
• Using a logged-in personal account. Your own history shapes the answer. Use fresh sessions.
• Changing prompts between rounds. Keep the list fixed so you can compare month to month. Add new prompts, but retain the originals.
• Ignoring the sources. The citations tell you exactly which pages the assistants trust. They are your to-do list.
• Chasing a single assistant. Customers use all three. A fix that helps one often helps the others, so measure all of them.
• Treating a missing mention as permanent. Results change as your content and third-party footprint grow.

Your Monthly Check Routine

Once the first audit is done, maintenance is light. Set a recurring calendar slot on the first Monday of each month.

1. Re-run your full prompt list in fresh sessions across all three assistants.
2. Record the new scores beside the previous month's.
3. Add 3–5 new prompts based on questions customers asked you that month.
4. Review the sources column for new sites that appear and old ones that disappear.
5. Pick the two weakest categories and schedule content to address them.

This takes about an hour a month when two people share the work. The record builds into a useful history: you can see which content changes moved the numbers and which did not.

What Good Looks Like

Set realistic targets. For most small businesses, a strong first-year goal is:

• Mentioned accurately in 100% of direct brand prompts
• Mentioned in at least 40% of location prompts
• Appearing in at least 25% of category prompts
• Your own domain cited in at least one assistant for each core service

These are working targets, not industry benchmarks. Your field, your location, and the number of competing providers all change what is achievable. Use your own baseline as the comparison point and aim to beat it each quarter.

Checking AI Recommendations FAQ

How often should I check whether ChatGPT, Gemini, and Perplexity recommend my business?

Run the full check monthly. Answers vary between runs, so a monthly cycle with the same 15–30 prompts gives you a stable trend. Check more often only when you have just published major changes and want early signs of movement.

Can I pay to be recommended by ChatGPT, Gemini, or Perplexity?

You cannot buy a place in an assistant's organic recommendations. Visibility comes from clear content on your own site, consistent business details, and mentions on trusted third-party sites. Improving those is the reliable route to being named.

Why does my business appear in one assistant but not the others?

Each assistant draws on different data and retrieval methods. Perplexity leans heavily on live web pages and shows its citations. Gemini reflects Google-related business data. ChatGPT combines learned knowledge with web lookups. A gap in one usually points to a specific weakness, such as an incomplete business profile or thin service pages.

What should I do if an assistant gives wrong information about my business?

Find the source of the error, using Perplexity's citations or a follow-up question about sources in ChatGPT. Correct the information on your own site first, then update the directory or profile that carries the mistake. Re-test after 2–4 weeks to confirm the change.

How does WorkspaceCMS help me get recommended by AI assistants?

WorkspaceCMS gives you one dashboard to build service pages, write answer-style content with AI-assisted drafting, check pages with built-in SEO scoring, and publish on a schedule through the content calendar. It does not run the assistant checks or guarantee recommendations. You run the audit, and WorkspaceCMS is where you publish the fixes.