Skip to content

Blog · Using AI

Can I just use ChatGPT to write my tender?

You can use ChatGPT on a UK tender, but it's unlikely to win: it misses what the buyer scores, loses track of long packs and invents details.

You can, but it's unlikely to win you the contract. ChatGPT doesn't know how your tender is scored, it loses track of long tender packs, it tells you what you want to hear and it fills gaps with details that sound right. In a tender, each of those costs marks, and some can cost you the bid.

Many small businesses try ChatGPT for tenders first, and it's a question we hear often. This guide explains where ChatGPT for tenders breaks down. It's based on interviews with a former public sector buyer and our beta users, OpenAI's own documentation and what we see when small businesses use TenderSeal.

Am I allowed to use ChatGPT on a tender?

Yes, unless the tender says otherwise. UK government guidance, PPN 017, says suppliers' use of AI "is not prohibited during the commercial process". Our guide Can I use AI to write a UK public sector tender? covers the rules. Being allowed to use it is one thing; whether it helps you win is what this guide is about.

Why doesn't ChatGPT work for tender writing?

ChatGPT is a good general writing tool. Tender writing asks for things it wasn't built to do: answer exactly what the buyer scores, with your own evidence, across every document in the pack, without inventing anything.

It doesn't know what wins marks

A buyer scores each answer against a scoring guide. ChatGPT doesn't know your scoring guide, the buyer's priorities or what your business has actually delivered, so it writes what a typical answer would say. The Scottish Government warns bidders that AI answers "tend to provide generic information", and that similar bids can result because the information AI draws on is available to other bidders too.

A former public sector buyer who now runs a bid consultancy told us how the AI drafts he has seen tend to score: "the aligning the response to the scoring guide isn't there yet. You know, it would probably score you a two or three out of five." Improving those drafts to a second or third version, he added, takes so long that it "is detrimental at the moment".

It loses track of long tender packs

A tender pack is rarely one document: the UK packs we checked ran to between 6 and 15 separate files, from instructions and specification to pricing schedules, forms and appendices. Your own evidence comes on top.

OpenAI's pricing page estimates the room for your input at about 12 pages on the free plan and about 40 on Plus in its Instant mode. Its Reasoning mode allows more on paid plans, but even then, longer input means less reliable answers: in tests of 18 models by Chroma, a company that builds AI search tools, accuracy fell as the input grew, even on simple tasks. OpenAI doesn't say what happens to the part of a long document that doesn't fit.

6–15

separate files in the UK tender packs we checked

TenderSeal's check of UK tender packs

12 pages

of input, OpenAI's estimate for the free plan

OpenAI pricing page, Instant mode

40 pages

of input, OpenAI's estimate for Plus

OpenAI pricing page, Instant mode

The result is a missed requirement: a mandatory form, a word limit, a pass/fail condition in an appendix. One of those can cost you the bid, however good the rest is.

It tells you what you want to hear

AI models are trained to be helpful, and research has found they accept the user's framing far more often than people do. On a tender, that means a confident yes.

It goes overboard and invents details

OpenAI's own help page says ChatGPT can produce incorrect or misleading outputs, including "fabricated quotes, studies, citations or references to non-existent sources", and advises: "Use ChatGPT as a first draft, not a final source."

In a bid, an invented detail becomes a certification you don't hold, a case study that never happened or a promise you can't deliver. Under the Procurement Act 2023 (England, Wales and Northern Ireland), a buyer can exclude a supplier that has acted improperly, including by providing information that is incomplete, inaccurate or misleading, where that puts it at an unfair advantage. The Act doesn't care whether a person or a tool wrote it.

Making it work becomes a second job

To keep ChatGPT on track, you have to feed it clean documents, write precise prompts, correct it question by question and check every line it writes. A beta user who bids on small tenders alongside her own business put it plainly:

“If you're not great with AI, you know, you do have to prompt it a lot … Whereas this thing, because you've uploaded all your documents and then it just generates the stuff, you know, I think that's a lot easier.”

Freelance copywriter, IrelandCall, 22 Jun 2026

In our experience, the back and forth often takes longer than writing the answer from scratch. And after all that, you have a generic draft that you still can't trust without checking every claim.

Here is what each problem costs you in a tender:

ProblemWhat it costs you
It doesn't know what wins marksGeneric answers that, in one former buyer's view, probably score two or three out of five
It loses track of long packsA missed form, word limit or pass/fail condition
It tells you what you want to hearA confident yes
It invents detailsA certification you don't hold, a case study that never happened, a risk of exclusion
Making it work is a second jobBack and forth that often takes longer than writing the answer yourself
The five problems, and what each costs you

Is ChatGPT safe for confidential tender documents?

Not by default. On ChatGPT's personal plans (Free, Plus and Pro), OpenAI says it may use your content to train its models unless you turn off "Improve the model for everyone" in your settings. Rating an answer with a thumbs up or down can make the whole conversation available for training, and even temporary chats may be kept for up to 30 days.

What does a tool built for tenders do differently?

It starts from the tender and your evidence, and it has guardrails built in, so you don't have to build them yourself on every question.

That's how we built TenderSeal, bid writing software for suppliers without a bid team. You give it your company context once, in three parts: your company profile (policies, insurance, turnover and certifications), your team (CVs) and your track record (past projects and proposals). Its Knowledge Hub learns from your past proposals, so each new tender starts from what you've already written. For each tender, TenderSeal:

  • reads the whole tender pack and pulls out the budget, requirements and evaluation criteria, with references to where in the tender each finding comes from
  • tells you if the tender is worth bidding on: a confidence score, your strengths and gaps, and a check of every eligibility requirement (how that check works)
  • splits the work into the answers you need to write, the forms you need to fill in and the documents you need to attach
  • drafts your answers from your own documents

You still review the draft before you submit, as you should with any bid. TenderSeal doesn't train AI models on your documents, and your data is encrypted.

Frequently asked questions

Will the buyer know I used ChatGPT?

Perhaps. Using AI isn't prohibited, but generic answers that don't meet the criteria lose marks, whoever wrote them. If the tender asks whether you used AI, answer honestly.

Can ChatGPT read a whole tender pack?

Not reliably. On the free plan, OpenAI estimates room for about 12 pages of input in its Instant mode, and even on paid plans accuracy falls as the input gets longer. A pack of several documents plus your own evidence is where requirements get missed.

Which AI is best for tender writing?

Claude, Gemini and Microsoft Copilot share ChatGPT's basic limits: they don't know your tender or your evidence unless you give it, and they can invent details. For tender writing, use a tool built for tenders, and check every claim before you submit.

ChatGPT's features, limits and plans change often; everything here was checked on 30 October 2026. This is general information, not legal advice.

Bring the tender. We’ll help with the rest.

TenderSeal reads the tender, answers your questions about it, and drafts your response to the scorecard.