Everything on Comms With AI so far has helped you make something. The library gives you the structure, Plan turns a need into a brief and a route, and the toolkits bundle the whole job. None of it told you whether the thing you made was any good. That changes this month: Message Test, the biggest addition to the platform since Plan, is live now and free to try. Alongside it, membership is open, and the tools directory has had its largest single update yet.
Message Test: a second pair of eyes before you send
Message Test does one job. Paste a draft, choose who is receiving it, and get a structured read on how it is likely to land.
You get seven things back: a one-sentence verdict, plainly stated with no score or grade; a short read on the likely reception; the three lines of what that audience actually hears, which is often not the message written; three objections they will raise, each with an answer you could actually give; the fact, date, figure or acknowledgement the draft needs and does not contain; a check on trust signals such as evidence, named accountability and specificity; and a tightened rewrite under 150 words, built only from the facts already in your draft.
There are eleven audiences to choose from, built around who actually receives comms work rather than who buys it, from trade journalist and sceptical frontline employee to regulator, funder and client approver. Or describe your own in a sentence. The same draft against two audiences is where most of the value sits: a redundancy announcement read as an employee, then as the line manager who has to defend it tomorrow, produces two different sets of problems.
Two limits, stated plainly. It is an assessment of how an audience of that type is likely to receive your draft, drawing on communications practice. It is not research, and it does not predict what any particular real person will think or do. And it will not help a message mislead the people reading it. Nothing you paste in is stored.
Guests get two tests a day. A free account gets five, no card needed. The full introduction, including a worked example on a restructure announcement, is in the launch update.
Membership is open
For teams that want Message Test and Plan without a daily limit, membership is now open at £20 a month, on fair use. The first 20 members keep 20% off for as long as they stay, and the discount applies itself at checkout. See what is included. The free library stays free, as it always has.
Tools directory: ten additions, and a hard look at synthetic audiences
The tools directory had its biggest single update this month: ten new entries, taking it to 74 tools across six categories.
Three fill gaps readers kept pointing out. Perplexity is finally listed: it answers questions by searching the live web and citing its sources, so every claim can be followed back before it reaches a briefing note. NotebookLM answers only from the documents you give it, which makes it the right tool for the papers comms teams actually work from. And Google Gemini joins the general assistants, notable less for capability than placement: if you run Google Workspace, it is already inside Gmail, Docs and Meet.
Two serve internal comms and planning: Workshop, an employee comms platform whose assistant was built for internal comms rather than adapted from a general model, and Simpplr Comms AI, which aims at the coordination layer, turning a campaign goal into a plan, audiences, channels and timelines in one workspace.
The remaining five map a space every comms team will be sold something from this year: audience simulation. Minds builds reusable AI personas from your own material and pretests messages against them. Listen Labs is listed alongside them deliberately as the honest counterweight: its respondents are real people, recruited at scale, with AI doing the moderating and the analysis. Three more sit on the watchlist while the claims settle: Synthetic Users, Evidenza and Aaru. The reviews say what each is actually good for, what remains unproven, and where a simulated answer should never substitute for asking your real stakeholders. Message Test lives near this space and draws the same line: assessment against an audience type is a second pair of eyes, not research.
From Applied / Comms With AI
Applied published one piece in August, and it pairs well with the directory update. “Five layers of AI tooling, and the comms job each one is for” (18 August) argues that the problem is not that comms professionals are using AI badly, it is that most are using one layer of it and assuming that is all there is, then maps the five layers and the comms job each one suits.
And fresh this week: “Lessons from OpenAI’s Hugging Face ‘break-in’: why orchestration is your next must-have comms AI skill” (3 September). Hundreds of AI agents probed Hugging Face’s systems without once asking a human for help, and Ethan Mollick’s response to it points at a future of work built on directing fleets of agents. The piece makes the case that the skills that future demands, briefing, judgement, coordination and quality control, are the ones comms people already have.
The ladder, from free to done-with-you
Everything above except membership is free and needs no registration. When a team wants to go further, the route is the same one every time. Start with the free Comms AI Readiness Diagnostic, a five-minute scored self-assessment that points each weakness at the right templates and the right next step. From there, a Consult session is a short, paid conversation with a written advisory note; Deploy is the productised build of your foundation documents, AI use policy, governance and tool instructions; and Manage keeps all of that current over time.