How a FishIntel Report Is Made

What data goes into a FishIntel report?

Ten layers of government and satellite ocean data, fetched at the same time for your zone's own water. Most layers have more than one source behind them, and the report says which one answered. Every source, with its update cadence and its own disclaimer, is named on the data sources page.

On the paid plan there is one more pass. An AI vision model reads the rendered satellite charts the way a captain reads them, as pictures, directed with our algorithm to find the temperature breaks, color lines, eddies and filaments that hold bait and fish. Those shapes are easier to see in an image than in a grid of numbers. What it finds is used to explain the ranked spots, not to add or move one.

Observations always win over models. Where no station reported a field, a model fills that one value and the report marks it as modeled.

How does FishIntel pick the spots?

The engine scores the whole of your zone on every ocean layer at once, keeps the water where the strongest signs overlap, and publishes the best three places as ranked spots, each with a measured run and bearing from a named landmark, a depth, a position for your plotter and a confidence. The analyst explains those spots; it does not choose them. A spot sits on a structure in the water, a temperature front, a current edge or a bend in sea height, and the color, salinity and plankton layers back it up, because a green patch is where bait might be and a seam is where it stays.

Every spot is inside your zone and on fishable water, never on land, in a lake or across a line into water that needs a permit you may not hold. Since we cover the Bahamas, if your zone is centered there a spot on the Bahamian side of the line is kept and marked. The spots are spread out, so three pins never describe one feature. Each region knows the depth band where its edges usually set up, and a spot outside that band is still published when the signs are strong, with the report saying it sits deeper than the usual water. A far spot has to earn its place over a nearer one, so you are not sent 40 miles for water that is only slightly better. And the run is measured, never estimated.

The map, the spot cards and the words always point at the same water. If the analyst's reasoning and the ranked list ever disagree, the list wins and the prose is corrected to match before you see it.

How is a spot's confidence set?

Confidence says how many independent signs agree on the same water and how fresh they are. It is set from the data, not by the analyst.

• HIGH: more than one structure agrees on the water, at least one supporting layer backs them, and the overlap covers real area rather than a single point.
• MODERATE: the evidence is thinner, or one of the layers behind the spot is older than its own age limit. A spot that would be HIGH on fresh data is demoted, not hidden.
• LOW: one sign carries it. The report still publishes a LOW spot when it is the best water available, and says so plainly, because an honest third choice is more useful than an empty one.

The score beside a spot (0 to 100) is how strongly the layers overlap there. Confidence is a different question: not how strong, but how sure. A strong overlap on old color can be MODERATE; a modest overlap that three fresh layers agree on can be HIGH.

How is the day rating set?

The day rating is about the boat ride, not the fishing. The engine reads the seas, the wave period, the wind and any marine advisory for the fishing hours and rates the day EXCELLENT, GOOD, FAIR or POOR, taking the worse of seas and wind. Any marine advisory or warning in force makes the day POOR, whatever the numbers say. The reason line under the rating names the numbers that set it.

A couple of important notes on the day rating. It is not a safety call: nothing in a FishIntel report tells you it is safe to go, and the National Weather Service forecast and advisories are what you check before you leave the dock, every time. And it is never a reason to pick or avoid a fishing spot: conditions live in the rating and the marine forecast, the fishing advice lives with the fish, and you weigh the two against your own boat.

How is the species outlook built?

Which species a report can name, the water temperature each one is fished in, the major current, the structure that holds fish and the landmarks the report may use all belong to a region, chosen from your zone's coordinates. Every region is built to a written fishery standard, with each species sourced and dated, and that standard, not any dataset, is what turns a coordinate into a species list.

For each species the engine compares your zone's own water temperature with that fish's band and hands the analyst a verdict: in the sweet spot, tolerable on the warm or cool side, outside it, or not surface-driven at all. Fish that live below the thermocline get no surface band, because a surface temperature says nothing about where they feed. The analyst turns the verdict into one reason and one tactic for today's water; it never invents a season, a limit or a temperature window.

Seasons and rules. Each species carries the season, size and bag rules we hold for that water, drawn from the state and federal regulators, sourced, dated, and re-checked every week. Where that calendar covers a fish, whether it is open or closed on the report day is worked out by the engine from the calendar, never by the analyst. Rules change, so the report tells you to confirm current regulations before you keep a fish, and links the agency for your water.

Water outside every modeled region is not refused and is not handed a borrowed species list. It gets the full set of ocean, weather and tide conditions, names no species, and says so when you pin it and in every report.

How is the report written?

The engine owns the numbers and the analyst owns the reasoning. The algorithm checks stand between the analyst and you. The analyst is handed everything above at once, the data, the ranked spots, the species verdicts, the day rating, the chart read, and is trained to do the one thing numbers cannot: say why bait should hold on that seam, which side of the spot to work, what to pull and how deep, and what the honest answer is when a layer shows nothing to fish. It writes in the voice of a working captain. In the Keys that voice is built on Captain Kit Carson's water; elsewhere it is an analyst who knows the region.

The analyst is held to a few rules that protect you. It never moves a spot or adds one. It never claims a season, a limit or a timing the data did not hand it. It never uses sea state, wind or an advisory as a reason to fish or avoid a place, that is each captain's decision. And it treats the charted Gulf Stream edge as what it is, a dated and approximate line, placing the Stream at the reef only when today's temperature and current agree.

Every sentence is checked. Before a report is stored, each claim the analyst makes is read against the data it was given: the tide table, the advisories in force, the wind and current directions, the forecast, the distances. A claim the data does not support is removed, and the removal is recorded with its reason, so a thin sentence is one that failed a check, not one that was never written. We regularly audit those records and constantly improve our algorithm.

A real morning: Alligator Reef, October 2, 2026

The sample report is this morning, with its maps. Here is how it came together.

The water. Satellite temperature climbed a soft two degrees from the reef toward a warm tongue to the southeast, no knife-edge break. Water color was thin that pass, pale green, no true blue in the frame. The current set northeast at under a knot inside the circle, with the real muscle of the Florida Current beyond the 50 mile ring. Salinity showed a broad slope, not a front. A Small Craft Advisory ran to 8 PM: east to southeast wind near 20 knots, 3 to 4 ft at 5 seconds.

The spots. Zone 1, an 11.4 nautical mile run off Alligator Reef in 239 ft of water: a temperature break, an eddy signal and a color front on the same water, HIGH confidence. Zone 2, 30.8 miles to the south: a current edge, an eddy and a salinity gradient, HIGH, but deeper than the usual band and on the Bahamian side of the line, so marked for a permit. Zone 3, 40.5 miles to the southwest: an eddy and the salinity gradient and nothing else, LOW. The nearer spot led because the far ones were not clearly better.

The rating. POOR. A Small Craft Advisory was in force for the Straits, and an advisory alone makes the day POOR whatever the seas and wind numbers say. The reason line under the rating names the advisory and shows what the numbers alone would have said, so the reader knows which one made the call.

The call. The analyst wrote that Zone 1 was the play: the green water off the reef holds the bait, the warming step pins it to a seam, the bend in sea height marks where the water turns, so troll the cooler, greener northwest edge of Zone 1 for mahi near the surface and run a planer for blackfin. Zone 2 was the alternative, a flow seam rather than a color seam. Every sentence of that summary passed the checks.

What a reader did with it. Read the rating first and decided about the ride. Read Zone 1 and its reason. Checked the chart ages in the data note (all inside their limits that morning). Then held the 5 second chop against his own boat. The two guide pages, How to read your report and Get the most out of FishIntel, walk through every part of the report.

What FishIntel gets wrong, and what it does about it

A fishing report is not one age; it is ten ages wearing the same date. A daily temperature analysis lands about a day behind. Ocean color can be days old under cloud. Satellite salinity is slow by nature. A buoy can go quiet and still hand you its last reading as though nothing happened. So every layer carries when its observation was taken, how old that makes it, which source answered, and whether that age is past the layer's own limit. The data note on every report lists the ages, and the dashboard charts are stamped. If you check our water temperature against your own sounder and find it two days old, we would rather you had read that on the report first.

Four things a layer can be, each said out loud:

The dangerous failure in ocean data is not the one that errors. It is the dataset that freezes and keeps answering: a retired feed will return a complete, healthy-looking grid forever, and nothing in that response says the numbers stopped being current months ago. We know because it happened to us in 2026, and three rules came out of it. A backup on the same host or satellite is not a backup; where we can, the standby is different infrastructure from a different producer. Backups are watched as closely as primaries, by an automated sweep that reads the publication date of every source and alerts when one falls behind. And a missing layer beats a stale one: a frozen dataset is never kept as a last resort.

Known weak points, stated rather than hidden:

• Productivity has one source and no verified backup, so cloud or an outage leaves it missing. Chlorophyll never stands in for it.
• Satellite color under cloud is often a day or more old, and the report says so and leans on the chart read and the temperature seam.
• Modeled values. Where no station reported, a model fills that value and the report marks it modeled. A zone with no live buoy still gets a report, on modeled conditions, and says so.
• Drafted regions. Outside the Keys the species bands have not yet been checked against a season of local water. Each report says so.
• The chart read is a model looking at a picture. It is evidence for interpreting a spot, never the thing that places one, and when it disagrees with the numbers, the numbers win.
• The analyst can still be wrong in a way the checks do not catch: a reason that reads well and is thin. Every report keeps a record of what was checked and what was removed, and the weekly audit reads it.

We publish this list because a report you cannot check is a report you should not trust. Every number in a FishIntel report says where it came from and how old it is, every gap says it is a gap, and when we find a weakness we write it here before a reader finds it on the water. That is the standard we hold ourselves to, and it is the reason the report is worth reading on the mornings it tells you the data is thin.

FishIntel is an analysis of ocean conditions. It is not a guarantee of fish and it is not a safety product. The providers' own disclaimers are quoted in full on the data sources page.

Common questions about how FishIntel works

Is FishIntel an AI tool that analyzes SST charts for offshore fishing?

Yes, and that is only one part of it. FishIntel reads NOAA and satellite sea surface temperature data every morning and finds the temperature breaks in your zone, so you do not have to interpret a raw chart yourself. Temperature is one of ten layers. The report also reads water color and productivity, salinity, currents and sea height, the buoys, the marine forecast and the tides, scores where those layers stack up, and has an AI analyst explain the strongest places in plain English, with a working captain's regional knowledge underneath it all.

Where does the data come from?

FishIntel uses NOAA and satellite data to provide its daily fishing reports. The sources are NOAA and other government and European satellite programs, NOAA buoys, the National Weather Service, and NOAA tide stations. Every source is named, with how often it updates, on the data sources page.

How does FishIntel decide where to put a spot?

Our own engine scores every cell of water in your zone on how many favorable factors overlap there and how strong each one is based on our algorithm. An AI analyst then explains the strongest places in plain English. The analyst never places a pin; the data does.

What does the confidence label mean?

HIGH means the report rests on fresh observations from several agreeing layers. MODERATE means something is older or missing but the picture still holds. LOW means too much is old or absent to trust a spot without checking it yourself. The data note under every report says which layers drove the call.

Does the AI ever make something up?

It is not allowed to. Every number in the report comes from a data layer, and a set of automated checks removes any sentence that does not match the data behind it. What the analyst adds is interpretation: why a seam matters, how to work it, what to watch for.

Why does my report say a layer is stale or missing?

Because it is, and we would rather you knew. Cloud blocks ocean color. Satellite salinity cannot read shallow water. A buoy can go quiet. The report says which layer, how old it is, and what it leaned on instead.

Does FishIntel use government data for its fishing forecasts?

Yes, FishIntel uses NOAA and satellite data to provide every report. Each one is built on NOAA data alongside other public satellite programs, documented on the data sources page.

Does FishIntel show species forecasts?

Yes. Each report names the species in season for your region, with the water temperature band each one is fished in and whether that band has been calibrated against local water. A zone outside every modeled region gets full conditions and no species list, and says so.

Does FishIntel tell me what is open or closed?

Yes, as a guide. Each species card carries the season, size and bag rules we hold for your water, drawn from the state and federal regulators, dated, and re-checked every week, and where that calendar covers a fish the report says whether its season is open or closed on the report day. Rules change faster than any calendar, so the report also tells you to confirm current regulations before you keep a fish and links the agency for your water. FishIntel is a guide to the rules, not the source of record.

Is FishIntel a safety product?

No. Read the National Weather Service forecast and current advisories before you leave the dock, every time, no matter what any report says about the fishing. A report can tell you what the water is doing; it cannot tell you what your boat, your crew and your experience can handle. The decision to go, and the responsibility for it, stays with the captain.