ChatGPT for Small Business: Choosing Your Setup
Table of Contents
Last verified: 17 September 2026, against OpenAI’s published pricing and plan comparison pages. Plan names, prices and limits in this category change several times a year, so the verification date matters more than the copy around it.
In short
Choosing ChatGPT for small business use is a decision about three things: who owns the account, whether your conversations are used to train models by default, and whether staff need to share tools and company knowledge with each other. OpenAI sells individual plans (Free, Go, Plus and Pro) and business plans (Business and Enterprise), and the dividing line between them is ownership and data handling rather than raw capability.
A sole trader or a single marketer is usually served by an individual paid plan. Two or more people using AI on customer, financial or staff data belong in a Business workspace, where OpenAI states that business data is not used to train its models by default and an admin controls the accounts. Cost is rarely the deciding factor, because a Business Standard seat and an individual Plus subscription sit at the same published monthly figure.
What does ChatGPT do well for a small business, and what does it not do?
It drafts, summarises, explains and restructures text faster than a person can, and it does none of your thinking about strategy. That’s the whole trade in one sentence.
ChatGPT is a large language model produced by OpenAI. It holds a conversation, reads long documents, drafts copy, answers questions in plain language and works through data you give it. For a business with no spare hours, that behaves like a quick assistant with no fixed working day.
What it does not do is decide which customers are worth chasing, audit your site, or know your market. It works from patterns, and it amplifies whoever is prompting it. Owners who get value from it are the ones who treat it as a drafting and thinking aid with a person still signing off the output, which is also the reason structured AI training for your team tends to pay back faster than another subscription does.
One more limit worth stating plainly. The model does not know what happened this week unless it searches, and it does not know your internal context unless you give it. Anything time sensitive, regulated or numerical needs checking against a real source before it leaves your building.
How should you choose a ChatGPT setup?
Judge the options against five criteria, in this order. Price comes last, because the paid tiers sit close together.
- Account ownership. If three staff use three personal logins, you control none of them. OpenAI states that Free, Go and Plus plans are designed for individuals, and Business plans start at two users. When someone leaves, their chat history, their custom tools and their prompts leave too.
- Default data handling. On individual plans, OpenAI’s plan comparison lists content as used to train its models with an opt-out available. On Business and Enterprise, it lists no training on business data by default. That single row is the main reason UK firms move.
- Sharing. Custom GPTs can be created from Go upwards, but sharing them across a workspace sits on the business plans, with Pro the single individual exception. Company knowledge and workspace agents are listed against Business and Enterprise only.
- Usage headroom. Heavy daily users hit limits. OpenAI now sells that headroom separately through the Pro tiers for individuals and a Premium seat inside Business.
- Admin and compliance. Single sign-on, an admin console, domain verification and SOC 2 Type 2 come with Business. SCIM provisioning, role based access controls, enterprise key management and data residency, including a UK region, are listed against Enterprise only.
Which ChatGPT setup fits which situation?
The table below maps each plan to the situation it answers, with the detail that usually settles the decision. Prices are OpenAI’s published list figures, set in US dollars.
| Setup | Who it suits | Published price | Training on your content | The detail that usually decides it |
|---|---|---|---|---|
| Free | Testing prompts before committing | $0 | Used by default, opt-out available | No custom GPT creation, limited uploads and limited deep research |
| Go | Light personal use, longer conversations | $8 per month in the US | Used by default, opt-out available | Custom GPTs can be created, but the advanced reasoning models are not included, and OpenAI notes the plan may include ads |
| Plus | A sole trader or one marketer using AI daily | $20 per month, billed monthly | Used by default, opt-out available | First individual tier with the advanced reasoning models, plus record mode, Sites, expanded deep research and expanded memory |
| Pro | One person running AI work for hours a day | From $100 per month | Used by default, opt-out available | Sold in two usage tiers, $100 and $200, with new sign-ups to the $200 tier paused since 10 September 2026. Bought for headroom rather than features, and worth it only after limits have actually stopped work |
| Business, Standard seat | Teams of two upwards handling customer or financial data | $20 per seat per month billed annually, $25 billed monthly | Not used to train models by default | Admin console, single sign-on, shared GPTs and company knowledge, for teams of 2 to 200 |
| Business, Premium seat | The two or three heaviest users in that team | $100 per seat per month billed annually, $125 billed monthly | Not used to train models by default | Five times the standard usage with no five hour limit, and seat types can be mixed in one workspace |
| Enterprise | Regulated or larger organisations | Custom, by quote | Not used to train models by default | SCIM, role based access, enterprise key management and data residency across ten regions |
Sources: OpenAI business pricing for Business and Enterprise figures and features, and OpenAI’s individual pricing page for Free, Go, Plus and Pro. Both checked on 17 September 2026.
What does ChatGPT cost a UK business?
Budget for the dollar list price plus VAT, and check the figure at checkout rather than trusting any guide, including this one. OpenAI publishes prices in US dollars per user per month. UK buyers are billed in a supported local currency with UK VAT applied at checkout, so the number on your invoice will not match the headline exactly, and it moves with the exchange rate. VAT registered businesses reclaim that VAT in the normal way.
Two costing points that catch people out. Business Standard is billed annually at the lower rate, so the monthly convenience option costs more across a year. And charities should look before they pay: OpenAI states that nonprofits can access a discount of up to 75 per cent on Business or Enterprise through its nonprofit programme.
Arithmetic worth doing before you commit. A five person team on Standard seats, billed annually at the published rate, costs the same per head as five separate Plus subscriptions. You get the admin console, the shared tools and the default data position at no extra cost per person. That’s the calculation that moves most small firms, and it’s why the plan question is rarely about money.
What actually changes when you move off personal accounts?
Three things change, and only one of them is visible on day one.
Control changes first. An admin adds and removes people, billing sits in one place, and work does not walk out of the door with a leaver. Data handling changes second: OpenAI lists business data as excluded from model training by default on Business and Enterprise, while individual plans carry an opt-out you have to remember to set. Sharing changes third, and it’s the one people underestimate. A custom GPT built on a personal Plus account helps one person. The same tool shared across a workspace, sitting alongside company knowledge, changes how a whole team answers a repeated question.
For anything touching customer records, staff data or financial detail, read the Information Commissioner’s Office guidance on AI and data protection before you roll anything out. The ICO notes that this guidance is under review following changes made by the Data (Use and Access) Act, so treat it as a live document and check the date on the page you are reading.
Where does ChatGPT save a small business real time?
Drafting, summarising and repetitive replies, in that order.
Content is the fastest return for most small firms. Blog outlines, product descriptions, newsletter drafts, FAQ answers and social captions all start faster with a machine draft. The draft is where it stops. The model does not know your voice, your local market or what your customers search for, so it belongs inside a content plan rather than in place of one, which is the argument for treating AI as one input into how your marketing work gets done rather than a separate exercise.
Customer communication is the quieter win. Templated replies for complaints, order updates and enquiries stop your team writing from scratch every time. That is a different job from putting an AI chatbot on your website, which brings its own handover, tone and UK GDPR decisions with it.
Internal work gets overlooked. Summarising a long supplier contract, turning a messy spreadsheet into a readable summary, drafting a process note, explaining a regulation in everyday English: these save hours nobody was counting. Anything time sensitive still needs verifying against the original source.
Should you build a custom GPT?
Build one when the same question is answered by different people in different words. That’s the honest test.
A custom GPT is a configured version of ChatGPT loaded with your own documents and rules. A professional services firm might load contract templates and standard clauses. A retailer might load the catalogue and the returns policy. Staff then get consistent answers without writing a detailed prompt each time. Creation is available from the Go plan upwards, and sharing one across a workspace requires a business plan or the higher Pro tier, which is another reason the setup question comes before the tooling question.
No coding is involved. The work is in deciding what the tool must refuse to do, what it should cite, and who checks it. That’s the step most firms skip, and it’s the step that decides whether anyone uses the thing in month two.
How does AI search change what your website needs to do?
It raises the value of content that a machine can lift a clean answer out of.
Assistants and AI overviews now answer questions that once sent a visitor to a website. Academic work on generative engine optimisation, published at KDD 2024 by a team led from Princeton, tested a set of content changes across thousands of queries and found that adding citations, quotations and relevant statistics improved how often and how prominently a source appeared in generated answers, while keyword stuffing tactics carried over from older SEO did not (Aggarwal et al., GEO: Generative Engine Optimization). Clear structure, named sources, direct answers and plain HTML are doing the work. Content hidden inside tabs, accordions or scripts is harder to read, because crawlers read raw HTML.
On llms.txt, the file some people suggest adding to signal your content to assistants: large scale testing has found no validated link between the file and how often a site gets cited. It costs little to publish and it’s a footnote, not a strategy. The structural work described above matters considerably more, and it sits inside your wider digital strategy rather than beside it.
Ciaran Connolly, founder of ProfileTree, puts the choice in blunt terms: “Most of the ChatGPT questions we get from owners are budget questions wearing a technical costume. The useful version is narrower. What work are you handing over, who else needs to see it, and what happens to that data once it is typed in? Answer those three and the plan picks itself. Skip them and you end up paying for seats nobody logs into.”
What does a sensible first month look like?
Start with one job, one person and one measure.
Pick a task that repeats weekly and currently takes too long: enquiry replies, meeting notes, product copy. Give it to one person on a paid individual plan for two weeks and have them keep the prompts that worked. If two or more people then want in, or the work touches customer data, move to a Business workspace before you scale rather than after. Add a written rule on what must never be pasted in, and a named person who checks output before it is published or sent.
The short version
ChatGPT for small business comes down to setup, not product choice. One person working alone is served by an individual paid plan. Two or more people, or any work touching customer and financial data, belongs in a Business workspace, where OpenAI states that business data is not used to train its models by default and an admin holds the accounts. A Business Standard seat and an individual Plus subscription sit at the same published monthly price, so the choice is about ownership, sharing and data handling. Every figure on this page was checked against OpenAI’s own pricing pages on 17 September 2026, and this category changes often enough that you should check the live page before you buy.
Frequently Asked Questions
Which ChatGPT plan should a small business start with?
A single user testing what AI can do should start on Plus, which is the cheapest individual tier carrying the advanced reasoning models, custom GPT creation and file uploads. A team of two or more, or anyone handling customer records, should start on a Business Standard seat instead, because the account ownership and default data handling are different rather than simply better.
Is the free plan good enough for business use?
It’s good enough for testing prompts and getting a feel for the tool. It isn’t a business setup: OpenAI’s plan comparison lists the free tier without custom GPT creation, with limited uploads and limited deep research, and with content used to train models unless you opt out.
How much does ChatGPT cost in the UK once VAT is added?
OpenAI sets prices in US dollars per user per month and bills UK customers in a supported local currency with UK VAT applied at checkout, so your invoice figure depends on the exchange rate on the day. Budget the list price plus 20 per cent and reclaim the VAT if you are registered, then confirm the exact number on the checkout page rather than any published guide.
Is it safe to put business data into ChatGPT?
It depends entirely on which account it is typed into. On individual plans, content is used for model training unless you opt out, so a personal account is the wrong home for customer, staff or financial data; on Business and Enterprise, OpenAI lists business data as excluded from training by default. Read the ICO guidance on AI and data protection before you set a policy, and write down what must never be pasted in.
What is a custom GPT, and do I need a paid plan to build one?
A custom GPT is a version of ChatGPT loaded with your own documents and instructions, so staff get consistent answers without writing long prompts. Creation is available from Go upwards, while sharing one across a workspace requires a business plan or the higher Pro tier, and the company knowledge feature behind it is listed on Business and Enterprise only.
Can ChatGPT replace a content or marketing role?
No. It compresses the drafting stage, which is often the most visible part of the job and rarely the most valuable one. Deciding what to publish, who it is for, what the search data says and whether a claim is defensible still needs a person, and unchecked output is exactly the kind of thin material that AI search now filters out.
I keep hearing I should be using ChatGPT for my business, but nobody tells me what for. What do I actually do with it on a Monday morning?
Take the task you put off every week because it is slow and boring, and hand that one over first: enquiry replies, a meeting summary, rewriting a product description, turning a page of notes into a client update. Work on it for two weeks, keep the prompts that produced something usable, and judge the tool on time saved rather than on novelty. If nothing on that list gets faster, the honest answer is that you don’t need a subscription yet.
Am I being upsold on ChatGPT Business, or does a five person company genuinely need it?
A five person company does not need it for capability reasons, because the underlying models are the same as on Plus. It needs it for ownership and data reasons: one bill, an admin who can remove a leaver, shared tools, and a default position where your conversations are not used to train the model. At the published annual rate a Standard seat costs the same per head as an individual Plus subscription, so there is little upsell in the arithmetic.
Do I need an llms.txt file for AI assistants to find my site?
No. Large scale testing has found no validated link between publishing the file and how often a site is cited by assistants, so it belongs in the footnotes rather than on a roadmap. Structuring pages so an assistant can lift a clean, sourced answer out of plain HTML does considerably more.
Getting your team using it properly
The plan is the easy part. Getting a team to use AI consistently, on the right jobs, with someone checking the output, is where the value sits, and it’s the part most firms leave to chance. ProfileTree works with SMEs across Northern Ireland, Ireland and the UK on exactly that, usually starting with a short audit of where AI saves time and where it quietly creates rework, followed by practical digital training for the people who will use it daily.