PredictWind's AI Weather Router or Your Own GRIBs? A Decision, Not a Verdict

On 21 October 2026, PredictWind launched AI Weather Router, a feature inside its Weather Routing and Departure Planning tools. The pitch is a written briefing, in the manner of a professional weather router, that tells you which departure day looks best and why.

Planning a passage? Get the free checklists and worksheet: join the newsletter and we'll send the Pre-Departure Checklists, the Catamaran Buyer's Guide and the Buyer's Worksheet straight to your inbox.

Our short answer: don't subscribe for the AI Weather Router's briefing alone. If you already use PredictWind routing, try the briefing alongside your own reading of the models. If you don't, start with free GRIBs, learn to compare models, and add routing once you can name what you are missing. We have not tested the feature, so until independent hands-on reviews appear, treat its go or no-go as one input to the skipper's decision, not the decision.

That raises a fair question for anyone planning a long passage: pay for a router's go or no-go, keep reading the models yourself, or do both? This is a decision piece, not a verdict. The only evidence so far is PredictWind's own launch release and feature pages. We have found no independent review, and we have not tested the feature at sea or anywhere else. Treat everything below about what it does as PredictWind's claim.

What PredictWind says it does

From the release and PredictWind's feature pages:

• It does not make its own forecast. PredictWind's routing algorithm first calculates routes using your boat's performance data and preferences, and the AI then interprets those results.
• It compares up to four departure windows. It can recommend a day, advise proceeding with caution, or say none of the options is fully recommended, with scores out of 10 for individual factors.
• It weighs wind, passage duration, upwind sailing, motoring, rain and potential thunderstorm activity.
• It breaks the passage into time periods and flags where forecast confidence drops, which PredictWind says is typically beyond about five days.
• PredictWind says it is "designed to support the skipper's judgement, not replace it."
• It is available to Standard and Professional subscribers in the PredictWind app, the Forecast website and the Offshore App. Boat-motion analysis, covering roll, vertical acceleration and slamming, is Professional only.

One detail we could not reconcile: the release says the briefing interprets routes across eight forecast models, while PredictWind's Weather Routing page still says you can compare the route from each of six models. We do not know which is current.

What any router assumes

A router is only as good as what you feed it, and that is the first failure mode.

Your polars. PredictWind's help pages say you must set up boat polars before routing. You can pick a predefined polar for a similar boat, enter simple sail polars, or import advanced polars. With the simple version you enter boat speed at 50°, 90° and 160° true wind angle in an average 15-knot wind, and PredictWind assumes that is the boat's maximum speed, with speed falling as the wind drops. A cruising boat with a heavy load, a dirty bottom or a cautious crew will not match a generic polar. Every route and every departure comparison inherits that error.

Your comfort limits. PredictWind's default comfort thresholds avoid roll above 4°, vertical acceleration above 0.2 g and slamming incidence above 50%, and its own help page says those defaults may not be comfortable. If the router cannot finish a route within those limits, it reruns the route without them so you still get a result. Read the output, not just the headline score.

The boat's dimensions. Motion modelling needs your boat type, displacement, waterline length, beam and hull draught. Wrong inputs give confident-looking nonsense.

Where a briefing can mislead

• Model disagreement can disappear in a summary. A tidy go or no-go is easier to follow than eight models arguing. When models disagree, that disagreement is the information. Open the underlying comparison before you trust the summary.
• Forecasts get less certain with time. PredictWind itself says confidence drops further into a passage. A score for departure day four of a ten-day passage is a guess about weather you will meet next week.
• A plain-language answer can sound more certain than the data. That is our worry about any written summary, not a finding about this product, which we have not tested.
• The release lists what it weighs. Local hazards such as a lee shore, a shipping lane, a bar crossing or a squall line that sits between grid points are not in that list. A router will not know your crew is tired.

When to override it

Override when the briefing says go and your boat or crew says no. Override when the fastest route puts you somewhere you cannot reef or heave to safely, or when it relies on a forecast period beyond the five days PredictWind flags. Override when your polar was a guess. And override when what you see on the sky and the barometer does not match the forecast. Departure planning is an input to the skipper's decision, not the decision.

What to still do by hand

• Compare at least two or three models yourself before you leave and note where they diverge.
• Read wind, gusts, sea state and swell period, not only a score.
• Check the forecast against what you observe at least daily underway.
• Write down your own limits, such as a wind and sea state you will not start in, before the briefing tempts you to argue.
• Keep a plan for when the files stop arriving.

Offline and satellite limits

PredictWind's help pages say weather downloaded through the Offshore App can be viewed offline, and that getting weather routes and GRIBs over a satellite connection requires a Standard or Professional subscription. The Offshore App connects to Iridium GO!, GO exec and Starlink, with an email-request option. PredictWind also says a weather routing file averages around 20 kB, far smaller than a GRIB.

The AI Weather Router is listed in the Offshore App. We could not confirm how large the written briefing is, or whether it works over a thin satellite link, so we would not assume it does. Whatever you use, download before you leave, and plan for stale forecasts if the link goes quiet. We wrote about layering comms in our Iridium piece.

Who should use it, and who should skip

It could suit you if you plan passages of several days or more, you want a structured second opinion on departure day, you are comfortable setting up polars and boat dimensions properly, and you will still read the models. It may also help a crew that wants a briefing everyone aboard can follow.

Skip it for now if you daysail or coast-hop and free forecasts answer your question, if you will not maintain accurate boat data, if you tend to follow a score instead of reading the weather, or if you are not yet sure a routing subscription pays for itself. PredictWind's free plan includes a GRIB viewer and 11 forecast models, so you can learn to compare models yourself without paying.

Prices as listed on PredictWind's pricing page on 25 October 2026: Standard is $249 USD a year and Professional is $499 USD a year, each with a seven-day free trial. The page also advertises a three-month satisfaction guarantee. We have not read the terms. Check the page before you decide, because prices and plans change.

Our call

We would not subscribe for the AI briefing alone. If you already use PredictWind routing, try the briefing next to your own reading and see whether it shows you something you missed. If you do not, start with free GRIBs and learn to compare models, then add routing once you can name what you are missing. Wait for independent hands-on reviews before treating its go and no-go as more than one input.

FAQ

Does PredictWind's AI Weather Router make its own forecast?

No. According to PredictWind, its routing algorithm first calculates routes from your boat's performance data and preferences, and the AI then interprets those results as a written briefing. Everything about what it does is PredictWind's claim; we have not tested it.

When should you override the briefing?

Override it when the briefing says go and your boat or crew says no, or when the route puts you somewhere you cannot reef or heave to safely. Also override it when the call leans on forecast periods beyond about five days, where PredictWind itself says confidence drops, when your polars were a guess, or when the sky and barometer don't match the forecast.

Can you learn to compare weather models without paying?

Yes. PredictWind's free plan includes a GRIB viewer and 11 forecast models, so you can practise comparing models yourself before deciding whether a routing subscription pays for itself.

Picking Your Departure Day From the Canaries: Read the First 72 Hours

If you decide to subscribe, you can use our PredictWind link. We have not tested the AI Weather Router, and the link does not change our view of it.

Disclosure: the PredictWind link above is an affiliate link. If you subscribe through it, we earn 25% of your first forecast subscription, hardware is outside the commission, and it costs you nothing extra. See our affiliate disclosure.

Charts, Checklists & Sea Stories — practical gear and passage notes for a family planning a circumnavigation. Free email when a new piece goes up.