StockTitan methodology
How Rhea-AI scores market news
Rhea-AI is StockTitan's proprietary news analysis engine. It reads every press release and SEC filing published on StockTitan, about 50,000 documents a month, and gives each one two independent scores:
- Impact: how far the stock is likely to move after the news, up or down
- Sentiment: the tone of the language
Version 8 rebuilt Impact around the move a trader actually sees once the news is out, pre-market and after-hours included, and on the same news it beats the engine it replaced on every accuracy measure we track. This page explains what the scores mean, how we test them, and where they stop.
The two scores
Sentiment measures tone. Impact measures the move.
Each score answers a different question, and they move independently. A press release can be neutral in tone and still send a stock up 30%, while enthusiastic language often comes with news the market barely notices. Reading the two together is what makes the feed quick to scan.
Rhea-AI Sentiment
Is the language of this document positive or negative?
Multi-layered natural language analysis of the full text, combining word-level and contextual reading. The engine understands sector-specific terminology: clinical vocabulary in a biotech release is not scored the way the same words would be in a retail earnings report.
- 5Very Positive. Language dominated by positive terms: growth, breakthrough, record results.
- 4Positive. Mostly positive language with some neutral terms.
- 3Neutral. Balanced or predominantly factual language.
- 2Negative. Mostly negative language with some neutral terms.
- 1Very Negative. Language dominated by negative terms: crisis, default, failure.
Rhea-AI Impact
How far could this news move the stock, in either direction?
Impact is our estimate of the largest move the stock makes after the news, counted from the last trade before publication to the close of the trading session that reacts to it. That is the same day when the news comes out before or during market hours, and the next session when it lands in the last hour of trading or after the close. Pre-market and after-hours trading count, and a spike counts in full even if it fades, since that is the move a trader could have caught. Impact is direction-blind: a failed trial and a buyout offer can both score 5.
- 5Very High, 20% or more. Buyouts, pivotal trial readouts, bankruptcy, deals that remake a company.
- 4High, 10% to 20%. Earnings that surprise, major partnerships, regulatory decisions.
- 3Moderate, 4% to 10%. Analyst rating changes, product launches, executive changes.
- 2Neutral, 2% to 4%. Minor updates, small contracts, routine announcements.
- 1Low, under 2%. Regular disclosures and administrative filings.
For example, say a company reports quarterly results at 4:05 p.m. Eastern time, after the close, and the last trade before the release was at $20.00. That evening the shares touch $23.40 in after-hours trading, the next morning they dip to $19.60, and they end the next session at $21.50. The window runs from the $20.00 trade to that next close, and the largest move inside it is the jump to $23.40, which is 17%, so the news counts as a level 4 even though the stock finished the next day only 7.5% higher.
The same announcement can earn a different level on two different stocks, because Impact also weighs how each stock usually trades and how it reacted to similar news before, and a contract that barely moves a large company can send a small one up 40%. Size explains most of it, since a $50 million contract is a rounding error for a company worth $500 billion and can be worth half of a company valued at $100 million. A small biotech also tends to ride on a single drug, so one trial result can decide much of what the company is worth, and with fewer shares changing hands on a normal day, the same rush of buyers or sellers moves its price far more than it moves a mega-cap’s, where many analysts follow every release and part of the news is usually expected before it comes out.
Is the tone positive or negative?
Question answeredHow far could this move the price?
Directional: positive to negative
DirectionDirection-blind: size of the move only
The document itself
HorizonFrom publication to the close of the session that reacts, pre-market and after-hours included
Overwhelmingly positive language
A score of 5 meansA move of 20% or more is likely
Inputs
What goes in, and what stays out
Sentiment reads only the document. Impact starts from the document too, then weighs it against how the stock itself trades and how it reacted to news before, because the same headline can move a small biotech 40% and a mega-cap 2%, and a score that ignored that would mislead you on both of them.
+ Considered by the scoring engine
- The full text of the news release or SEC filing, including the headline
- The kind of event described: earnings, financing, trial readout, M&A, corporate action
- Financial figures and magnitudes stated in the document
- For Impact, the stock's own trading behavior before the news and its past reactions to similar news
- For Impact, when the news came out: before the open, during trading, or after the close
− Excluded by design
- Macroeconomic conditions and broader market moves
- The company's name and reputation
- Social media chatter and analyst opinions
- Anything that happens after the news is published
- Any judgment on whether the claims in the document are true
A score that relied on a famous name or on the mood of social media would drift with hype and carry its bias from one article to the next.
Rhea-AI judges each stock by how it has actually traded and reacted to news, so a small company's news is judged by the same standard as a famous company's, and nothing the market does after publication can leak into the score.
Calibration
What an impact level means in practice
We hold our scores to one standard, what the market actually did next. The figures below cover 76,648 news articles from March to August 2026, each scored by a version of Rhea-AI 8 that had only seen earlier news, and each matched to the largest move of its primary ticker after publication, measured the way Impact defines it. Moves are absolute values, so direction is ignored.
| Impact level | Articles | Avg. largest move | Moved 10%+ | Moved 20%+ | Moved 30%+ |
|---|---|---|---|---|---|
| 1 · Low | 10,217 | 1.6% | 0.2% | < 0.1% | < 0.1% |
| 2 · Neutral | 19,712 | 3.2% | 1.3% | 0.2% | 0.1% |
| 3 · Moderate | 28,439 | 7.2% | 17.5% | 3.0% | 1.0% |
| 4 · High | 11,413 | 15.5% | 58.2% | 19.7% | 7.9% |
| 5 · Very High | 6,867 | 34.7% | 84.8% | 54.4% | 34.8% |
These figures describe what followed past news, and no single article is guaranteed to behave like the average, since liquidity, float and the day's market all play a part. The move is the largest one after publication, so a spike that faded before the close still counts.
Past calibration does not guarantee future results.
Reading the tone
What a sentiment score means in practice
Sentiment answers one question: how positive or negative is the language of this document? Every word carries weight in the reading. A release built around debt, defaults, and bankruptcy reads low; one built around record results and successful launches reads high. The score describes the text itself, never the stock.
Same engine, different words · illustrative excerpts
“…announced a restructuring of its outstanding debt, citing substantial doubt about its ability to continue as a going concern and the possibility of bankruptcy protection…”
“…reported record quarterly revenue, driven by the successful launch of its flagship platform and accelerating growth across every segment…”
Highlighted terms illustrate how wording pulls a score down or up. The engine weighs the full text, and the same word can weigh differently depending on sector and context; the exact vocabulary and weights are proprietary and are not published.
+ What the score tells you
- How positive or negative the language of the document is, on the 1–5 scale
- Which way the wording leans: growth, breakthrough, and record results pull it up; debt, default, and bankruptcy pull it down
- Tone read in context: sector-specific terms are weighed by what they mean in that industry
- A reading computed the same way for every document, so you can compare tone from one end of the feed to the other
− What it does not tell you
- Anything about stock performance: sentiment does not predict the direction or size of any price move
- How the market will react: the score is computed from the text alone, before and independently of any reaction
- Whether the claims are true: upbeat language scores as upbeat language regardless of substance
- How big the move could be, which is Impact's job, since the two scores move independently
Press releases are written by the companies they describe, and framing matters: most issuers present even routine news in the best possible light.
You should expect the sentiment distribution to lean toward the positive half of the scale, and it helps to read scores against that baseline, where a Neutral on a company-authored release is often closer to lukewarm and a Negative one means the language could not be framed upward.
Weighted takes
How much each point weighs
Every news item and SEC filing on StockTitan comes with Rhea-AI's positive and negative points, and since a small certification and a failed primary endpoint should never look equally important, each point now carries a weight, judged by the same rules for every company.
Major
A point that on its own changes the picture of the company's business or finances, such as a primary trial endpoint met or missed, guidance raised or cut, a going-concern doubt, an agreement to sell the company, or a financing, deal or loss that is large next to the company's market cap.
Moderate
A point that moves the numbers without changing the overall picture, such as growth or a decline in a core metric, a margin shift of a few points, a secondary endpoint, or a contract or cost that matters but is not large for this company.
Minor
Detail and context, such as a certification, an integration, a small grant, exploratory data, or a figure that repeats a point already weighted.
When the bars are hollow, the point is forward-looking, a plan, a forecast, a proposal or a term that depends on something still to come, so it has not happened yet and may not happen. The text of the point says so as well, with words such as expects, plans or proposed.
When a point is about a one-off dollar amount, like an offering, a loan, a deal or a charge, a small tag sets that amount against the company's market cap at the time Rhea-AI read the document (the last close available then). A US$2.5 million loan means little to a company worth billions and a great deal to one worth US$4.7 million, and the weights follow the same logic. We only put the tag on a Moderate or Major point. A Minor point with a dollar figure is often a number the document already counts somewhere else, like a charge that is part of a larger loss, and a percentage beside it would make it look more important than its weight says it is.
The bar above the points is plain arithmetic. A Minor point counts 1, a Moderate point 3 and a Major point 9, and each side's weights are added up, so a single Major point outweighs several Minor ones and a long list of small positives cannot hide one serious negative. When the positive side holds more than 65% of the total the bar reads Positive, when the negative side holds more than 65% it reads Negative, and anything in between reads Mixed, which is the shaded band in the middle of the bar.
When every point in a document is Minor we leave the bar out, since a verdict added up from small details alone would look just as firm as one resting on a large contract or a failed trial. The points are still listed with their marks, and we draw the bar whenever at least one point on either side is Moderate or Major.
Illustrative example
How this balance works
Rhea-AI gives every point it takes from this document a weight. Minor counts 1, Moderate 3 and Major 9, so one Major point outweighs several Minor ones. The bar adds up the weights on each side, and when neither side holds more than 65% of the total the balance reads Mixed.
It reads the document as published, with the same rules for every company, and it does not look at what the market expected or at how the stock traded, so a point can be objectively good on a day the stock falls.
Rhea-AI Sentiment measures something else, the tone of the wording.
Hollow bars mark forward-looking points.
Positive
- Moderate pointThe treatment was well tolerated, with no new safety signals.
- Minor pointSecondary measures trended in favor of the treatment.
- Minor pointPatients reported better daily function at nine months.
- Minor pointThe company received a small research grant.
- Minor pointEnrollment finished ahead of schedule.
- Minor point. Forward-looking: it has not happened yet and may not happen.The company plans to present the full data at a medical meeting.
Negative
- Major pointThe trial did not meet its primary endpoint.
- Major point. Forward-looking: it has not happened yet and may not happen.The company will review strategic options, including a sale.
- Moderate point. Forward-looking: it has not happened yet and may not happen.A proposed $6 million loan facility would fund operations. 30% of market cap
+ What it tells you
- How the points in this one document weigh against each other, by the same rules for every company, large or small
- Which points matter most, sorted first on each side, with minor details folded away on long lists
- Which points have already happened and which are still plans or forecasts
− What it does not tell you
- What the market expected. A company growing revenue 70% has reported a good result even if traders hoped for 90% and the stock falls on the day.
- Where the stock is going. The balance is not a forecast of the price.
- Whether the document is accurate. Rhea-AI weighs what the company disclosed, and the source document stays one click away.
Sentiment and the balance answer different questions. Rhea-AI Sentiment reads the tone of the wording, how positive or negative the language is, while the balance weighs what the document actually discloses, point by point. The two usually agree, and when they do not, the gap is often worth a closer look, as with a trial that missed its main goal described in upbeat language.
The weights are generated by Rhea-AI when it reads the document, and the explanation next to each bar shows when. If a point or a weight looks wrong to you, tell us through the contact page and we will review it.
Engine accuracy
Rhea-AI 8 against the engine it replaced
We scored the same 6,945 news articles from September 2026 with both engines, news that neither engine's development had ever used, and compared each score with the move that followed. Every figure below is how much better Rhea-AI 8 did than the previous engine on that same news.
Test sample: 6,945 news articles published from September 1 to 18, 2026, kept out of all development and scored once, with both engines reduced to their impact level on the same articles. On that news, Rhea-AI 8 also rated High or Very High 11% more of the news that was followed by a move of 10% or more, its High and Very High scores were followed by a move of 10% or more 15% more often, and it rated earnings reports that went on to move the stock 10% or more High or Very High 38% more often. A separate comparison on 23,613 articles from July and August 2026 found 63% fewer companies whose news all received the same impact level, a known weakness of the previous engine.
Validated cases
Impact 5 in the wild
Three articles from 2026, scored at publication by the Rhea-AI engine live at the time, with the market reaction measured afterward by StockTitan's own tracking. Each links to the original article, where the score and the measured reaction are still visible.
KORE announces agreement to be acquired by Searchlight Capital Partners and Abry Partners
Innodata reports record first quarter 2026 results
Quince Therapeutics announces topline results from pivotal Phase 3 NEAT clinical trial
The QNCX case shows why impact is direction-blind: a pivotal trial readout is a maximum-impact event whichever way it resolves. Impact told readers this document mattered before the market decided which way it mattered.
Validation
How we test a scoring model before you see it
Every generation of the scoring engine goes through the same discipline before it is allowed to score live news, and Rhea-AI 8 went through all of it.
Historical grounding
Rhea-AI 8 learned from hundreds of thousands of news events in StockTitan's own archive, each paired with the move that followed it, across NYSE, Nasdaq, and OTC listings.
Tested in time order
Candidates are tested month by month on news they have never seen, the way they would have run live, so results cannot benefit from hindsight.
A final exam
Before release, the engine was scored once on a stretch of recent news kept aside for that purpose and compared with the engine in production on exactly the same news. The accuracy figures on this page come from that exam.
Release and monitoring
The new engine was timed on the server that scores live news before it replaced the old one, and after release its calibration against real market reactions stays under continuous monitoring.
The exact inputs, weights, and architecture of the scoring engine are proprietary and are not published.
What we publish is everything needed to judge the scores from the outside: their definitions, what they consider and leave out, their measured calibration against real market reactions, and their limitations.
Material changes to the engine are announced in the version history.
Limitations
What Rhea-AI cannot tell you
Reading the scores well means knowing their boundaries. These are the ones that matter.
- Not investment advice. Scores are statistical indicators built to speed up research, not recommendations to buy or sell.
- No claim verification. Rhea-AI scores what a document says, not whether it is true. A confident press release from an unreliable issuer can still score well.
- Market-wide swings. Impact reads the news and the stock, not the whole market, so on days when everything moves together a stock can move more or less than its level suggests.
- High score, quiet stock. 3 in 100 Impact 5 articles in our testing moved the stock 5% or less, and a low score does not rule out a surprise either.
- Thin trading. Impact counts every trade, pre-market and after-hours included, so on a thinly traded stock a handful of trades can set the size of the move.
- AI-generated text can err. Summaries and takes are generated automatically and can contain mistakes. The original source document is always one click away.
Accountability
Who is behind Rhea-AI
Rhea-AI is built and operated entirely in-house by the StockTitan engineering and editorial team. The scoring engine was developed with guidance from financial industry professionals, and its output is reviewed continuously against professional standards.
All AI-generated content on StockTitan is labeled as such, kept visually separate from source documents, and accompanied by its limitations. When the engine changes in a way that affects what readers see, the change is recorded in the version history on this page.
- About StockTitan · who we are and how the platform works
- Editorial policy · standards for accuracy, review, and corrections
- Version history · every material change to Rhea-AI since 2020
FAQ
Common questions
What does the Rhea-AI Impact score measure?
Impact estimates how far a stock is likely to move after a news release, up or down. The move is measured from the last trade before the news to the close of the trading session that reacts to it, with pre-market and after-hours trading included, and the largest move in that window counts even if it fades by the close.
Level 1 means under 2%, level 2 means 2% to 4%, level 3 means 4% to 10%, level 4 means 10% to 20%, and level 5 means 20% or more.
What is the difference between Rhea-AI Sentiment and Impact?
Sentiment measures the tone of the language in a document, from Very Negative to Very Positive.
Impact estimates how far the stock is likely to move after the news, regardless of direction.
They are independent: a neutral-tone document can carry Very High impact.
Does Rhea-AI Impact predict which way a stock will move?
No. Impact is direction-blind and estimates only the size of the move, so a level 5 can mean a 40% jump or a 40% drop.
Sentiment does not forecast direction either, since it measures the tone of the language, and company press releases lean optimistic. To judge direction, read the news itself and weigh what it says, using the Impact level as a guide to how much could be at stake.
Does the Rhea-AI Sentiment score predict stock performance?
No. Sentiment measures only the tone of the language in a document: wording about debt, defaults, or bankruptcy reads low, wording about record results and successful launches reads high.
It is computed from the text alone, before and independently of any market reaction, and says nothing about how the stock will perform.
Because press releases are written by the companies themselves, tone tends to lean optimistic, so scores should be read against that baseline.
What does a Rhea-AI Impact score of 5 mean in practice?
In testing on news from March to August 2026, news rated Impact 5 was followed by an average largest move of 34.7% in the primary ticker. More than half of those articles moved the stock 20% or more, and about one in three moved it 30% or more.
It is a strong historical association, and no single article is guaranteed to follow it: 3 in 100 Impact 5 articles moved the stock 5% or less.
Why did a stock rated Impact 1 jump 30%?
It happens, and it is rare. In testing on news from March to August 2026, 2 of 10,217 articles rated Impact 1 were followed by a move of 30% or more, and about 2 in 1,000 by a move of 10% or more.
When it does happen, look for something the article itself does not contain, such as a second piece of news later in the same session, a rumor spreading on social media, a swing across the whole market, or a stock so thinly traded that a few orders set the price. Impact scores what the article says at the moment it is published, so it cannot see what comes after.
How accurate is the Rhea-AI impact engine?
Compared with the previous engine on the same 6,945 news articles from September 2026, news neither engine's development had used, Rhea-AI 8 matched the exact impact level 13% more often and missed by two levels or more 26% less often.
It missed 70% fewer big moves, meaning news it rated Low or Neutral that was then followed by a move of 10% or more, and 15% more of its High and Very High ratings were followed by a move of 10% or more.
What data does Rhea-AI use to produce its scores?
Sentiment is computed from the text alone of each news article or SEC filing.
Impact starts from the same text, what kind of event it describes and what it says, and weighs it against how the stock itself usually trades and how it reacted to similar news in the past.
Neither score uses the company's name or reputation, social media, analyst opinions, or anything that happens after the news is published.
Is Rhea-AI Impact the same for OTC stocks?
Yes. OTC stocks get the same definition and the same five levels as NYSE and Nasdaq stocks, with pre-market and after-hours trades counted in the same way.
OTC news tends to be followed by larger moves, so OTC stocks receive High and Very High levels more often. In testing on news from March to August 2026, 37 in 100 OTC news items were followed by a move of 10% or more, against 29 in 100 on Nasdaq and 11 in 100 on NYSE.
OTC shares also trade more thinly, so a few large orders can set the size of a move, and the exact level is a little harder to call on OTC stocks than on listed ones.
Does a high score guarantee accuracy or predict stock movement?
No. A high score does not validate claim accuracy, assess company credibility, or guarantee that a stock will move.
Rhea-AI outputs are algorithmic, statistical indicators, calibrated against historical market reactions.
How is Rhea-AI validated?
Every candidate engine is tested month by month on news it has never seen, in time order, so hindsight cannot help it.
Before release, the engine was scored once on a stretch of recent news kept aside for that purpose and compared with the engine in production on exactly the same news, and it was timed on the server that scores live news.
After release, calibration against measured market reactions is monitored continuously and published on this page.
How does Rhea-AI weigh the positive and negative points?
Each point gets a weight, Minor, Moderate or Major, judged by the same rules for every company and measured against the company's size.
The balance bar adds them up with Minor counting 1, Moderate 3 and Major 9, and reads Mixed when neither side holds more than 65% of the total.
It reads the document as published and does not consider market expectations or the stock price.
We leave the bar out when every point is Minor, and only a Moderate or Major point gets the small tag that sets a one-off dollar amount against the market cap.
Should I make decisions based solely on Rhea-AI?
No. Rhea-AI is a research aid, not financial advice.
Always consider company fundamentals, broader market conditions, and your personal risk tolerance, and consult a licensed advisor where appropriate, before trading or investing.
Update cadence
Version history
Every material change to the engine since launch is listed here, and continuous monitoring runs between the releases.
- Impact now measures the move after the news, from the last trade before publication to the close of the session that reacts, with pre-market and after-hours trading included
- On the same news, 13% more articles land on exactly the right impact level than with version 7, and 70% fewer big moves are missed (news rated Low or Neutral that is then followed by a move of 10% or more)
- 63% fewer companies whose news all received the same impact level
- Complete rebuild of the impact engine, verified on a held-out month before promotion
- New extreme-move alert and materiality signal for outsized movers
- Scoring latency reduced to milliseconds per article at publication
- Overhaul of the sentiment engine with context-aware scoring of industry-specific language
- Greatly improved accuracy for biotech, pharmaceutical, and healthcare news
- Smarter recognition of corporate actions such as offerings, reverse splits, and regulatory decisions
See the scores on live market news
Every article on StockTitan carries its Rhea-AI scores the moment it is published, next to the original source document.
Open the live news feed About StockTitanRhea-AI outputs are automated, statistical indicators provided for informational purposes only and do not constitute investment advice, an offer, or a solicitation. Sentiment is computed from document content alone; Impact also considers the stock's own trading history. Neither score verifies the accuracy of any claim. Market data on this page reflects StockTitan's own measurements for the periods stated; past results do not guarantee future outcomes.