ChatGPT
Which accounting software should a company with 10 people use?
- Ledger: clear plans for teams of 10.
- Books: step-by-step help for the first month.
- Sums: plain prices for small teams.
How to get AI to recommend your company
Published Updated 13 min read
Our numbers: OpenAI model (gpt-5.6-luna) via Codex CLI with web search, not the ChatGPT app, 01-02.11.2026, 13 software questions.
ChatGPT mostly names companies from pages it finds online. You cannot apply or pay for a place in the answer. Nobody can promise one, Nenn included.
Card needed. Cancel before day 7 and nothing is charged.
- ledger.example/pricingPricing page
- help.books.example/first-monthHelp page
- sums.example/for-teams-of-10Page for one kind of customer
The short answer
To get ChatGPT to recommend your company, make the pages it reads easy to quote. There is no form and no paid place in the answer.
An OpenAI model, 115 runs, 366 citations. The method is in the section on how we got these numbers.
Key points
No sign-up. You cannot sign up or pay to be named in a ChatGPT answer.
It searches. For most buying questions, the model searched the web and named companies from what it read.
Prices first. Pricing pages got 40 % of citations. Help pages got 14 %. Reddit got 0.
Then check. Make those pages clear, let OpenAI's search bot in, and check the answers every week.
How does AI pick the companies it names?
For most buying questions, ChatGPT searches the web, reads a few pages and names the companies those pages support.
Which accounting software should a company with 10 people use?
ChatGPT
software for 10 people
site:ledger.example
books help first month
ledger.example/pricing
books.example/help
sums.example/teams
- 1. Ledger1
- 2. Books2
- 3. Sums3
It searches
Often several web searches per question
It reads
A few pages from the results
It names
Companies those pages support
It links
Some of the pages it used
Scheme: the question leads to 3 web searches, 3 pages are read, 3 companies are named, each with a link.
See how often the model searched the web, question by question
Why your own pages are the part you control
Being recommended by ChatGPT means your company is named in its answer to a buying question. Nobody hands out those places. The model builds each answer from pages it finds and reads at that moment. So the part you control is your own pages. If they answer the question in plain words, the model can use them.
Some answers skip the search. 10 of our 115 runs answered without searching. 9 of those 10 came from one short 'pick one of two' question. Without a search, the model answers from what it learned earlier, so a new page cannot help there yet.
Which pages did AI cite in our runs?
Pricing pages most of all, about 4 in 10 citations. Reddit got no citations at all.
Which pages got the 366 citations
- 148Pricing and plan pages
- 50Help and docs pages
- 20Blog posts and guides
- 11Review sites and lists
- 0Reddit
- 137All other pages, such as product, feature and home pages
See the full table and how we counted
We asked an OpenAI model 13 buying questions about software, 115 times, with web search on. Its answers carried 366 citations. We sorted every cited page by words in its address. Each page counts in one group only.
| Page type | Citations | Share |
|---|---|---|
| Pricing and plan pages | 148 | 40 % |
| Help and docs pages | 50 | 14 % |
| Blog posts and guides | 20 | 5 % |
| Review sites and lists | 11 | 3 % |
| 0 | 0 % | |
| All other pages, such as product, feature and home pages | 137 | 37 % |
13 questions, 115 runs, 366 citations, sorted by words in the page address (for example /pricing, help., /blog). Each link counts once, in the first group that fits, in the order of the table. One question asked for a price directly, which adds to the pricing count. OpenAI model (gpt-5.6-luna) via Codex CLI with web search, not the ChatGPT app, 01-02.11.2026.
What does AI look for on your own website?
Mostly your prices. When the model searched inside one website, about half of those searches asked for prices.
site:ledger.examplepricing
128 of 258
site: searches that looked for prices
One clear pricing page
Give your prices one clear page with a plain address, such as /pricing.
Prices as text
Write plan names and prices as text, not inside an image.
What is a site: search?
A site: search looks inside one website only, like 'site:ledger.example pricing'.
In our runs the model made 258 site: searches. 128 of them looked for prices.
6 steps to be the page AI finds
Start with the pages AI cited most in our runs. None of these steps is a trick, and none comes with a promise.
- 1
Put your prices on one clear page.
Name each plan, say what it includes and for how many people, and give the price as text. Pricing pages got 148 of 366 citations in our runs.
- 2
Answer one question per help page.
Open with one sentence that answers it, then add the details. Help and docs pages got 50 citations, second only to pricing. A Peec test (August 2026) found a direct first sentence to be the strongest single edit.
- 3
Write a page for each kind of customer.
12 of our 13 questions named a team size or a trade, like 'a law office with 5 lawyers'. Pages made for one trade or team size got 63 citations. Use those words in a page title and first lines.
- ledger.example/for-law-offices
- books.example/for-bakeries
- sums.example/for-teams-of-10
- 4
Use your blog for real questions, with your own facts.
Blog posts and guides got only 20 citations in our runs, so do this after steps 1 to 3. One question per post, with numbers you can show.
- books.example/blog/first-month-checklist
- 5
Let OpenAI's search bot read your site.
Allow OAI-SearchBot in your robots.txt file. OpenAI says sites that block it are not shown in ChatGPT search answers.
User-agent: OAI-SearchBot Allow: / - 6
Check the answers every week.
The same question can bring a different list each time, so one look tells you little. You can check by hand first.
See which pages AI cites for your customers' questions. Start 7-day trial
How does Nenn help you become the page AI finds?
Nenn shows which pages AI cites for your customers' questions, then drafts the page you are missing.
Sources
14 cited domains · 3 engines · Last 7 days
Citation share0.0%
Own citations0
Cited domains14
Your rank-of 14
#DomainCategoryShareCitationsEngines
- 1.ledger.example15
- 2.books.example12
- 3.sums.example9
- 4.software-reviews.example7
- 5.accounting-guide.example6
- 6.office-forum.example4
- -your-site.exampleYours0Not cited
Is it worth the work if nobody can promise a mention?
It is, when you start with pages your buyers need anyway. A clear pricing page and good help pages serve people first.
Buyers want the same facts
A buyer who lands on your pricing page wants the same facts AI wants: plan, price, what is included. So the work pays off with people even in a week when AI names someone else.
Start small
Pick the one question that brings you the most customers and fix that page first.
What should you avoid when you try to get named by AI?
Skip fake reviews, self-ranked 'best' lists and anyone selling a place in the answer.
Do
Do: Keep your name, prices and address the same on your site, profiles and directories.
Do: Ask real customers for honest reviews.
Do: Publish your own data with a method line.
Do not
Avoid: Buy or write fake reviews.
Avoid: Publish a 'best tools' list with yourself at number 1.
Avoid: Fill pages with lists of keywords.
Avoid: Pay anyone who promises a place in ChatGPT answers.
ChatGPT seldom cites self-ranked lists. Peec studied 232,000 citations about software from December 2025 to February 2026. About 3.6 % of ChatGPT's citations went to such lists. Peec still advises against them.
How did we get these numbers?
We asked the questions ourselves and counted every citation. Here is the setup and what it cannot tell you.
See the full setup and its limits
- Engine
- An OpenAI model (gpt-5.6-luna) through Codex CLI with web search. This is not the ChatGPT app, and not your account.
- When
- 1 to 2 November 2026.
- Questions
- 13 questions a company might ask before it buys software, each asked 5 or 10 times in a fresh chat.
- What we counted
- Every cited link in the answers, and every search query the model wrote.
- Limits
- One engine, one field (software), one model version. Pages are sorted by words in their address, not by reading them. A different field or engine can give other numbers.
Questions about getting recommended by AI
Is there a form to add my company to ChatGPT?
No. ChatGPT finds companies through the pages it reads on the web. Your own site is the part you can change today.
Can I pay for a place in ChatGPT's answers?
No. ChatGPT shows ads on some plans, but they are marked as sponsored and sit apart from the answer. OpenAI says ads do not change what ChatGPT answers.
Does posting on Reddit help ChatGPT recommend me?
In our runs, 0 of 366 citations went to Reddit. Pricing pages got the most. Other studies found other results. A large Peec study ranked Reddit the most cited site across five AI engines, ChatGPT included.
How long after I change a page can ChatGPT use it?
Nobody can give a date. OpenAI says a robots.txt change takes about 24 hours to reach its search. Whether a page then gets cited depends on the question and on the other pages it finds.
Do I need special code or an AI file on my site?
No special file is needed. For ChatGPT, let OAI-SearchBot read your site (step 5). Google says its AI Overviews and AI Mode need no special files, markup or schema. Clear text on a page that bots can reach matters more.
Do these steps also help with Perplexity and Google AI?
The steps are general, but our numbers come from an OpenAI model only. Each engine picks sources in its own way. So check each engine on its own.
See which pages AI cites for your customers' questions
Each week Nenn puts your customers' questions to three AI engines and drafts the pages you are missing. 3 plans, each with a 7-day trial.
Card needed. Cancel before day 7 and nothing is charged.
Which accounting software should a company with 10 people use?
- 1Ledger
- 2Books
- 3Your company?
Keep reading
Three next steps, from the other article to the tools that run the checks for you.