Consilyn Start free

Can you use AI on an expert network call? What the rules say

Guide ยท Updated

AI tools can now transcribe a call and suggest what to say next. For an expert taking paid consultations, the question is not whether this works but whether you are allowed to use it, and what you owe the client if you do.

This guide covers what several large expert networks' published terms say, how recording consent works, and what happens to confidentiality when call text goes to an AI service. It is general information, not legal advice. Terms change, so read your own network's current terms before relying on anything here. The pages below were read on 3 October 2026.

Can you use AI on an expert network call?

Often not. Several large networks forbid or restrict AI-generated content in consultations, and two of those below forbid experts recording or transcribing calls themselves, one naming AI tools expressly. Where your network forbids it, do not use a live tool on its calls unless it gives you written permission. Where it is allowed, the client still has to agree.

The client's agreement does not override a network's ban, and a network's permission does not replace the client's agreement.

Is preparing with AI before a call different from using it during the call?

Yes, in ways that matter. Preparing on your own, from your own material, captures no one's voice, so recording consent does not arise. But some networks' AI rules cover preparation too.

Take an invented example. An expert in hospital purchasing is booked for a call about how health systems negotiate supply contracts. Beforehand, she drafts notes from her own experience and asks an AI assistant to tidy them into short points. No call audio is involved, although some networks' rules may still forbid the AI step. During the call, a live tool would have to capture the client's voice and usually send the words to a service that writes suggestions, and that is where consent law, network rules and confidentiality all apply.

Two cautions about preparation. First, the client's screening questions can be confidential in their own right: some network terms list the questions a client asks among the information you must protect, so pasting them into an AI chatbot sends them to a third party. Second, some AI rules are not limited to the live call. One code forbids using AI even to assist in answering client questions. Another forbids using or relying on AI-generated content in consultations and in written responses to project opportunities, which on a plain reading covers screening answers and answers prepared in advance to read out. If in doubt, ask the network.

What do large expert networks' terms and codes say?

The pages we read differ in wording but point the same way: give your own first-hand view, do not present AI output as your own, keep the client's identity and questions confidential, and in some cases do not record or transcribe calls at all. These are summaries and general information, not legal advice; follow the links and read the current versions.

GLG

GLG's Network Member Code of Conduct has a section on AI-generated content. Apart from AI that GLG itself builds into its systems, members "may not use, or rely upon, AI-generated content" in phone consultations, surveys, written work or written responses to GLG project opportunities. The same code says members must not disclose a project or a client's identity to any third party.

GLG's member terms, linked from the code and marked last updated April 2022, include an undertaking that you will not record or transcribe your consultations with clients, alongside your consent to clients recording and transcribing them. Members sign the terms each year, so check the version you last signed. GLG's own preparation advice describes calls as a discussion of your direct knowledge and experience, and says there is typically no need to prepare anything in advance.

Guidepoint

Guidepoint's advisor terms and conditions, last updated 4 February 2026, are the most direct on recording. Advisors may not record or transcribe consultations "through the use of AI tools, personal devices, or other methods", though some clients may. The terms also forbid misrepresenting your experience or knowledge through AI or otherwise, and treat client names, the questions a client asks and the topics discussed as confidential information.

AlphaSights

AlphaSights' Expert Code of Conduct deals with AI under plagiarism. Experts must not use AI to produce insights and pass them off as their own, and using information directly and exclusively from AI tools such as chatbots is prohibited. The plagiarism rule it sits under covers vetting as well as calls. The code also requires every engagement to stay confidential, including its participants, its subject matter and the client's identity. It does not address recording; it refers to separate Terms of Engagement.

Third Bridge

Third Bridge's Expert Code of Conduct is the broadest. Experts must not use any form of AI, such as large language models, "to answer or assist in answering any questions" from Third Bridge or its clients. Its Expert Terms and Conditions, last revised December 2024, require you to keep confidential the client's identity, the subject of the consultation and the questions asked.

Other networks have their own rules, and all of them can change; Guidepoint's terms, for example, say changes take effect when posted.

You should get it, and in many places the law requires it. Live transcription works by capturing the other person's voice, which is the kind of activity call-recording and interception laws deal with. Some places require one party's consent, others everyone's, and on a call across borders the stricter rule may apply. This is general information, not legal advice.

In the United States, federal law sets a minimum of one party's consent, and states can be stricter. The Reporters Committee for Freedom of the Press, in an introduction to recording laws written for journalists, lists a number of states that primarily require the consent of everyone in the conversation, including California, Florida, Illinois, Pennsylvania and Washington. When participants are in different states, courts have differed over which state's law applies, so the cautious course is to assume the stricter one does.

Outside the US the rules vary again. In the UK and the European Union, data protection law generally applies to recording or transcribing what identifiable people say, and other countries have their own requirements.

Two points are easy to miss. A tool that never saves audio still has to capture it to transcribe, so do not assume it falls outside a recording law. And the law is only the floor: a network's terms can forbid what the law allows.

What happens to confidentiality when call text goes to an AI service?

It leaves your control. The words are processed by another company under that company's terms, which may allow it to keep the text, use it to improve its products or have people review it. Since the network terms above treat the client's identity and questions as confidential, you need to be sure you are entitled to share them.

Read the provider's data terms before the first call, not after, because free and paid tiers often differ. Google's Gemini API terms, for example, say that on unpaid services Google uses submitted content to improve its products, that human reviewers may read it, and that you should not submit sensitive, confidential or personal information there. On paid services, the same terms say prompts are not used to improve Google's products, though Google logs them for a limited period to detect misuse. Users in the UK, the European Economic Area and Switzerland get the paid-service data terms either way.

The client's identity needs particular care, because a transcript can reveal it through introductions and context even if no one names the firm. Keep the client's name and firm out of any notes the tool keeps. And some material should never reach any tool because it should not be on the call at all, such as non-public information about a current or former employer.

What should you check before using AI on a call?

Work through these in order and stop at the first one that says no. The first four carry most of the weight.

  1. Read your network's current terms and code of conduct yourself, not a summary such as this one. Look for recording, transcription, AI and confidentiality.
  2. If the network forbids live tools or recording, do not use one on its calls. If you think there is a case for an exception, ask the network for written permission and keep the reply.
  3. Check the consent rules where you and the client are. If you cannot tell, assume everyone on the call must agree.
  4. Tell the client at the start of the call, in plain words, and go ahead only if they agree.
  5. Keep confidential material out: the client's name and firm, non-public information about any employer, and anything price-sensitive about a listed company.
  6. Read the AI provider's data terms, including whether text is kept, used for training or read by people, and choose a tier that suits client work.
  7. Prefer tools that do not save audio and that clearly label AI-written text, so you never present a model's general knowledge as your own experience.
  8. Answer from your own experience. Whatever appears on screen, what you say is your statement, and the client is paying for your view.

How should you tell the client?

Briefly, at the start, before any substance, and in a way that makes it easy to say no. Say what the tool does, which company receives the text, whether that company may use it, and that audio is not saved if that is true. Then ask, and accept the answer without fuss. If your tool asks you to confirm what you told the client, say at least that.

For example, with a free Google Gemini key: "Before we start, I'd like to mention that I use an assistant on my laptop that transcribes our conversation as we go. It never saves the audio, and it sends the text to Google's AI service to suggest answers, mainly from my own notes. Google may use that text to improve its products. Are you comfortable with that, or would you prefer I switch it off?"

On a paid tier whose provider does not train on the text, say that instead: "...sends the text to a paid AI service, which does not use it for training, to suggest answers."

If the client hesitates or says no, switch it off and carry on; it may reflect their firm's own rules rather than anything about you. And only ask where your network allows the tool in the first place.

How Consilyn fits

Consilyn is a live answer screen for experts, in your browser on Windows and Mac. Used on a call your network allows, with the client's agreement, it transcribes the call on your laptop without saving the audio, and shows short read-aloud answers built from your own screening answers and background, each labelled with its source. Every AI-written answer is marked not verified. Where it draws on your own material, a green label names the source and your exact words are quoted underneath. It will not start listening until you confirm you have told the client the call is transcribed on your laptop and its text sent to Google's AI, and they are happy to go ahead. It can also prepare answers to likely questions before the call, which involves no call audio, though network rules on AI still apply.

The call text and your material go to Google Gemini using your own key. On Gemini's free tier Google may use that text to improve its products, and people may review it. Google's terms advise against sending confidential information to the free tier, so consider a paid tier for client work. Consilyn does not decide whether a call allows it: your network's terms and your client do.

Check your network's terms first

Where live help is allowed and your client agrees, Consilyn shows short answers built from your own screening answers, each labelled with its source. It asks you to confirm what you told the client before it listens, and audio never leaves your laptop.

Start free in your browser