Search interest data measures how much attention a topic receives, not whether investors are bullish or bearish. A spike in searches for "recession" or "Bitcoin" shows rising relative interest, and it does not show that anyone is buying, selling, fearful or optimistic. Because the link between a search and an investment view is weak, search data works best as a low-confidence context layer that is checked against market-based evidence such as volatility, credit spreads, breadth and positioning.

Search Interest and Narrative Intensity as Sentiment Data

By Swoopr Editorial Team Published October 1, 2026

Educational research written with AI assistance and reviewed under the editorial policy. Not investment advice.

What does search interest actually measure?

Search interest measures attention: how often people type a phrase into a search engine, compared with everything else they search for. It is a window into curiosity and worry, and a poor window into portfolios.

Google Trends is the best-known source. Google describes its data as a largely unfiltered sample of real search requests that has been anonymized (no one is personally identified), categorized (each query is assigned a topic) and aggregated (grouped together). Each data point is then divided by the total searches in the chosen region and time range, and the results are scaled to a range of 0 to 100. These properties come from Google's own explanation in Google Trends: FAQ about Trends data.

That design makes the series very good at one job, which is showing how interest in a topic changes relative to its own history. It makes the series unsuitable for another job, which is counting searches. A reader who remembers one idea from this page should remember this one: the number is a relative index, not a volume.

This page looks at search interest and narrative intensity as alternative data. The related page on social sentiment and retail activity signals covers a different family of evidence: what people post and trade on social platforms and brokerage apps. Search data records what people look up, which is a quieter and less performative behavior than posting.

How should you read the 0 to 100 scale?

The scale is relative, so three habits keep interpretation honest.

A value of 100 is a peak, not a total. It marks the highest relative interest inside the selected comparison. A value of 50 means interest at roughly half that peak. Neither number says how many people searched.

The comparison context shapes the number. Because values are measured against total searches in the selected region and time range, two charts of the same phrase can look different if the region or window differs. Google's help page on results by region states that popularity is relative to the total number of Google searches in a specific place and time, and that an unhighlighted region does not mean there is no interest. Two places can show the same index value while having very different absolute search counts.

Small numbers are fragile. Google states that low-volume terms appear as zero and that the data include statistical noise, so it is "not a perfect mirror of search activity." A one-day jump in an obscure query is more likely to be noise than a burst of public attention.

Should you use a search term or a topic?

Google Trends offers two kinds of object, and the choice changes what you are measuring.

A search term is literal: it tracks the exact words typed. A topic groups related searches that share one concept or refer to the same real-world entity, including different spellings and languages. Google's guidance in Compare search terms and topics says terms suit exact brand names or phrasings, while topics suit broad concepts and avoid spelling and language fragmentation.

For market questions the distinction matters in practical ways:

Company names. The word "apple" can mean a fruit, a company or a product. A company topic removes most of that ambiguity.
Macro concepts. "Recession" as an exact term answers a question about that specific word. A broader recession topic answers a question about general public interest in economic downturns.
Ticker symbols. Many tickers collide with ordinary words or acronyms. A finance-oriented topic, where one exists, is usually safer than a raw string.

A careful reader notes, for every chart, whether it shows a term or a topic, because a comparison between a term and a topic is a comparison between two different constructs.

Why attention is not direction

Attention has size but no sign. A surge in searches for "Bitcoin" can happen during a rally or a crash. Searches for a bank's name can rise because customers are worried, because journalists are covering a story, or because students are researching a case study. Interest in a company often spikes around earnings whether the report is good or bad.

Analysts try to recover direction by grouping queries into baskets.

A fear-oriented basket might include phrases such as recession, stock market crash, layoffs, bank failure and bear market.

An opportunity-oriented basket might include phrases such as how to invest, best ETF, bull market and IPO.

Even a well-built basket is a narrative proxy, not a behavioral fact. Someone searching "should I sell stocks" may be deciding to hold. The word in the query does not reveal the action taken afterward. That is why baskets are described here as measures of narrative intensity, which means how unusually much attention a theme is getting compared with its own past, and not as measures of fear or greed.

Google Trends can draw on different search surfaces, such as web search, news search, YouTube, shopping or image search where supported. A spike in news search for a central bank may reflect event-driven information seeking. A spike on a video platform for a speculative asset may reflect a different audience with a different motive. Mixing surfaces in one chart without labeling them blends unlike behaviors.

The "Trending now" feature is a separate product from the Explore charts. Google's page on Trending now describes it as showing queries with a recent surge that relate to a news story, refreshed on a short cycle, and notes that its chart is exact-match while the Explore chart is broad-match. A list of what is trending today therefore cannot be stitched onto a long Explore history as if both were one measurement.

Why do seasonal patterns create false narratives?

Many financial searches rise and fall with the calendar. Interest in tax filing, retirement account contribution limits, student loan rules or holiday shopping stocks tends to follow predictable cycles. A rise in such searches is expected, so it carries little information by itself.

The question to ask is whether interest is unusual for this time of year. Google's help page on comparing search interest over time offers quick comparisons against the preceding period or the same period in the prior cycle, which is a simple way to separate a one-off surge from a regular seasonal rhythm. More formal approaches include a year-over-year comparison, a same-week-of-year baseline, or a rolling median by calendar period.

Events, manipulation and audience bias

Event contamination. Many spikes trace to one event: an earnings release, a product launch, a fraud allegation, a regulatory action, a central bank decision or a sudden market move. The spike is real attention, but it is not a lasting sentiment regime. Annotating a chart with the event that caused the jump turns a mystery into an explanation.

Irregular activity. Google says it filters automated searches and queries that may be tied to attempts to manipulate results, but it also says the data are not a perfect mirror of search activity. Markets add their own incentives. FINRA and the SEC's Office of Investor Education and Advocacy have warned in Social Sentiment Investing Tools: Think Twice Before Trading Based on Social Media that such tools can be inaccurate, incomplete or misleading, that posts can be used to push prices, and that investors should not rely solely on them. That alert concerns social sentiment tools and not search data, but the lesson carries over: narrative activity can be amplified or manufactured, and it can drift away from fundamentals.

Audience bias. Search behavior varies by country, language, age and platform. Google users are not the same group as securities investors. Search data tends to capture people actively looking for information, and it captures professionals using terminals or proprietary research much less. It is therefore a reasonable gauge of public narrative and a weak gauge of professional positioning. For positioning evidence, see the page on Commitments of Traders net positions.

Where does search data sit among sentiment sources?

Sentiment evidence comes in families that differ in how close they sit to real behavior:

Market prices and spreads. These come from actual transactions and quotes. Examples are the Cboe Volatility Index and credit spread measures.
Regulatory and exchange data. These report real exposures but arrive with a delay, such as FINRA margin debt.
Surveys. These record stated opinions but carry sampling and response bias, as covered in the guide to investor surveys as contrarian signals.
Search and narrative data. These are timely and broad, but the step from "searched this phrase" to "holds this view" is the weakest of the four.

Search data is not discarded because of this ranking. It is weighted according to what it measures. The page on the market sentiment source ladder lays out the general approach to ranking sources by trust, and the broader research source ladder explains how evidence quality is graded across the site. Timing also matters, since a series that revises or arrives late is harder to compare. The page on data latency and vintages explains why the date a number was retrieved belongs next to the number.

A framework for narrative intensity

A practical reading of search data uses three layers, each answering a different question.

Layer 1: attention intensity. How unusual is current interest compared with its own history? A percentile rank over a stated window answers this. A reading in the 90th percentile means interest is higher than in about nine tenths of the sample period. This layer has no direction.

Layer 2: narrative family. Which theme is the attention about? Common families include macro fear, inflation and cost of living, recession and employment, market crash, speculative enthusiasm, technology enthusiasm, crypto interest, safe-haven interest and income interest. Grouping makes the chart readable, but every inclusion changes the construct, so the list of queries matters as much as the result.

Layer 3: cross-market confirmation. Does market-based evidence agree? Volatility, credit spreads, market breadth, positioning and leverage are checked against the narrative reading. The page on cross-asset risk appetite describes how to read several markets together, and the guide to combining sentiment with price action covers the same idea from the chart side.

Search interest rarely deserves to decide the picture alone. Its main value is as a prompt: when attention is extreme, it tells a careful reader which other evidence to examine.

Four combinations of narrative and market evidence

Comparing attention with market conditions produces four common situations. They describe what the data show together, not trades to make.

Situation	Attention	Market evidence	Reasonable reading
1	Fear-themed searches rising	Volatility up, credit spreads wider, breadth weaker	Public attention is consistent with a broad stress period
2	Fear-themed searches high	Volatility falling, credit stable, breadth improving	Narrative attention may be lagging market conditions
3	Enthusiasm-themed searches rising	Participation broadening, credit stable	Attention fits wider risk appetite, though crowding is still worth checking
4	Enthusiasm-themed searches rising	Participation narrowing, volatility rising	Excitement may be concentrated in a few names and fragile

The second and fourth rows are the most instructive, because they are the cases where attention and markets disagree. Disagreement is not a forecast. It is a reason to look at which evidence family is more direct, and the source ladder usually favors the market-based evidence.

How to build a narrative basket that can be trusted

A basket of queries is only useful if another person could rebuild it and get the same answer. Seven habits help.

Define the concept first. "Macro fear" might be defined as attention to recession, unemployment, banking stress and market-crash themes. The queries follow from the definition, not the reverse.
Prefer topics where available. They reduce language and spelling noise.
Record every inclusion and exclusion. Adding a phrase after it happened to work in hindsight is a form of overfitting, because the basket then describes the past instead of measuring something.
Label versions. A basket called "macro fear, version 1.2" tells a reader that earlier charts used a different list.
Normalize each component before combining. Separate Google Trends pulls are each scaled to their own peak, so raw 0 to 100 values from different pulls are not automatically comparable.
Take care when stitching long histories. Joining several retrieval windows requires overlapping periods or another documented rescaling method.
Limit the weight of the whole family. Search data should not dominate a composite of several sentiment families. The page on the sentiment composite framework discusses family weighting.

A reader evaluating someone else's search-based chart can use this list as a checklist: if the basket, the version, the region, the surface and the retrieval date are not stated, the chart cannot be reproduced.

A worked example with illustrative numbers

The figures below are invented to show the method, not real readings.

Suppose a basket of three macro-fear series has these percentile ranks over a five-year window: "recession" topic at the 92nd percentile, "stock market crash" term at the 88th, and "layoffs" topic at the 80th. The median of the three is the 88th percentile. A median is often more robust than an average when one series is distorted by a single news event.

Now suppose market evidence reads as follows: the volatility index at the 55th percentile of its own history, a high-yield credit spread at the 42nd percentile, equity breadth improving, and futures positioning near neutral.

A useful summary reads: "Public attention to macro fear is high. Market-based stress measures are moderate and participation is improving, so the narrative is elevated but not confirmed by higher-trust evidence." That statement is more accurate than calling the market "extreme fear," because it names which evidence is high and which is not.

Notice also what the example leaves out. It does not say what will happen next, and it does not say what to do. It describes a gap between attention and market behavior, and gaps like that can close in either direction.

What search data does not tell you

It does not show who is buying or selling. Transactions, flows and positioning data do that.
It does not show how much money is at risk. Leverage and exposure data, such as margin debt, speak to that.
It does not show whether the narrative is true. A heavily searched story can be accurate, exaggerated or false.
It does not provide a stable signal. Relationships between attention and prices change as platforms, audiences and news cycles change.
It does not replace participation measures. The difference between opinion and participation is explored in the page on sentiment versus breadth, and the page on positioning, sentiment and breadth compares the three directly.

Common mistakes

Treating 100 as a volume. It is a normalized peak in the chosen context.

Assuming more searches means more bullishness. Attention measures interest, and interest can be fear.

Comparing independently scaled series as if they share a unit. Each pull is scaled to its own maximum.

Ignoring the term versus topic choice. A literal term can be narrower and more ambiguous than a topic.

Reading low-volume spikes literally. Noise is a larger share of small series.

Ignoring seasonality. Calendar rhythms can look like sentiment events.

Rewriting the basket after seeing outcomes. That builds hindsight into the measure.

Overweighting alternative data. Search interest complements market-based evidence and does not replace it.

For the broader picture of how these indicators are built and misread, the overview of how sentiment indicators work is a good companion, and the macro and market regimes section shows the backdrop against which attention shifts. The market sentiment hub lists the full set of related guides, and the glossary defines terms used here.

Confidence is separate from intensity

A reading can be intense and still deserve little trust. Intensity says how high the attention is. Confidence says how much the observation can be believed.

Confidence rises when the query has enough volume, the topic mapping is unambiguous, the signal persists across several observations, several related queries agree, and the configuration has not changed. Confidence falls when a series is sparse, driven by a single event, drawn from a mismatched region, newly introduced, or inconsistent across repeated pulls.

The distinction prevents a common error. A one-day spike in a low-volume query can post a very high intensity reading with almost no confidence behind it. A moderate but persistent rise across several related topics may deserve more weight. Reporting both numbers side by side tells a reader how seriously to take the chart.

Frequently Asked Questions

Does Google Trends show the number of searches?

No. Google Trends shows relative interest, normalized by total searches in the selected region and time range and scaled from 0 to 100. A value of 100 marks the peak of interest within that comparison, and it says nothing about how many searches occurred. Two charts with the same value can represent very different absolute volumes.

Does a search spike mean investors are buying?

No. A spike shows attention, which can come from fear, excitement, news coverage or simple curiosity. It does not reveal whether anyone traded, in which direction, or in what size. Transaction and positioning data are more direct evidence of behavior, though they have their own delays and limits.

Should I use a search term or a topic?

Use the object that matches the question. Topics group related searches and variations in language and spelling, so they suit broad concepts or entities. Exact terms suit cases where the precise wording matters, such as a brand name or a particular phrase. Whichever you pick, note it beside the chart, because the two measure different things.

Can search data be used for market timing?

Search data is timely, but sampling, normalization, ambiguous intent, noise and shifting relationships make it unsuitable as a standalone timing tool. Any pattern that looked useful in the past may not repeat. It is better treated as context that raises questions, not as a trigger.

How much weight should search interest get in a sentiment composite?

As an alternative-data family it generally earns a lower weight than market-based or regulatory evidence, because the connection to actual behavior is weaker. The weight should be stated openly, capped, and tested for sensitivity. Tuning it to maximize past results is a way of fitting history, not of measuring sentiment.

What is narrative intensity?

Narrative intensity describes how unusually much attention a theme receives compared with its own history. It says nothing about whether the narrative is accurate or whether prices will move in any particular direction. It is a descriptive measure of how loud a story has become.

References

Swoopr Editorial Team

The Swoopr Editorial Team produces educational investment research and tools covering stocks, ETFs, bonds, crypto, and portfolio strategy. All content is reviewed for accuracy and adherence to our editorial policy.

Found an error? Report a correction.

Direct Answer