Google, Bing & YouTube Autocomplete Trend Monitor

Track how autocomplete suggestions change on Google, Bing, and YouTube

Monitor the exact keyword terms you care about and receive structured events only when autocomplete changes.

This Actor keeps a persistent baseline for every term × source. After the first successful observation, it can report:

• a suggestion that appeared;
• a suggestion that disappeared and stayed absent long enough to be confirmed;
• a suggestion whose position moved by a meaningful amount.

Google, Bing, and YouTube are intentionally treated as independent monitoring sources. A Google suggestion is never compared against a Bing or YouTube baseline.

No search volume is invented. No CPC is estimated. No keyword-difficulty score is fabricated. No browser automation or LLM is used in runtime.

Key features

• Google autocomplete monitoring
• Bing suggestion monitoring
• YouTube autocomplete monitoring
• Persistent term × source baselines
• NEW_KEYWORD events
• Confirmed LOST_KEYWORD events
• RANK_CHANGE events
• Configurable minimum rank movement
• Consecutive-miss confirmation before loss
• Google/YouTube language targeting
• Google/YouTube country targeting
• Source-specific state by construction
• One autocomplete request per term × source
• No alphabet expansion
• Free first-run baseline
• Free RUN_SUMMARY
• Health and source-contract checks
• Controlled pacing and retries
• Pay Per Event
• No fake search volume

Unofficial community Actor. Not affiliated with, sponsored by, or endorsed by Google, Microsoft/Bing, YouTube, or Alphabet. The Actor uses public autocomplete/suggestion routes. Source behavior, availability, ordering, and suggestions can change at any time.

---

What this Actor is for

A keyword suggestion extractor answers:

What autocomplete suggestions exist around this seed right now?

This Actor answers a different question:

What changed in autocomplete since I last checked?

That distinction makes it useful for:

• SEO monitoring;
• brand-query monitoring;
• content research;
• YouTube topic monitoring;
• competitor-term monitoring;
• product and category research;
• long-running keyword intelligence;
• trend observation;
• automated SEO dashboards;
• alerting and scheduled workflows.

---

Important: this is a monitor, not a keyword-expansion scraper

For every term, the Actor makes exactly one request per selected source.

It does not perform expansions such as:

term a
term b
term c
...
term z

or question/preposition permutations.

That keeps the product lightweight and makes the monitored unit clear:

one exact term
×
one autocomplete source

If you want broad keyword discovery, use the companion:

Google, Bing & YouTube Keyword Suggest Scraper

Use this Actor when you already know the terms you want to watch.

---

Important: autocomplete trend does not mean search volume

Autocomplete changes can be useful as an observable search-interface signal.

They are not direct search-volume measurements.

This Actor does not return or invent:

monthly search volume
CPC
keyword difficulty
paid difficulty
search intent score
Google Trends index

A RANK_CHANGE means the suggestion changed position in the autocomplete list returned by that source.

A NEW_KEYWORD means a suggestion was not present in the previous successful state and is present now.

A LOST_KEYWORD means it remained absent for the configured number of consecutive successful checks.

Nothing more is claimed.

---

First run creates a free baseline

The first successful read of each term × source unit creates the baseline.

Example:

term: coffee maker
source: google
current suggestions: 10
previous state: none

Result:

baseline created
0 NEW_KEYWORD
0 LOST_KEYWORD
0 RANK_CHANGE

The Actor does not present the current list as if every suggestion had just appeared.

From the next successful observation onward, the list is diffed against persistent state.

---

Source state is separated by construction

The same seed can look very different across engines.

For example:

coffee maker / google
coffee maker / bing
coffee maker / youtube

are three different monitored units.

Their state keys and event identities are source-specific.

That means a Google list can never be diffed against a Bing list.

This prevents cross-engine differences from becoming false change events.

---

Sources

Google

Google autocomplete uses the public suggestion route configured by the Actor.

For Google, the request includes:

hl
gl

using the configured language and country.

---

YouTube

YouTube suggestions use the same public suggestion family with the YouTube dataset flag.

The configured language and country are also applied.

---

Bing

Bing uses the public OSJSON suggestion route implemented by this Actor.

The route does not receive the configured language/country parameters.

Therefore:

language
country

affect Google and YouTube monitoring, not Bing.

---

Who it's for

SEO teams

Monitor strategic seed terms and watch for:

• emerging autocomplete suggestions;
• disappearing suggestions;
• meaningful position changes.

Content teams

Watch category or question seeds that matter to editorial planning.

Brand teams

Monitor autocomplete around:

• company names;
• products;
• executives;
• branded queries.

The Actor does not classify a suggestion as positive or negative.

YouTube creators and video teams

Track topic autocomplete for the exact YouTube search seeds you care about.

Market and competitor researchers

Watch terms related to:

• products;
• software categories;
• competitors;
• industries;
• consumer problems.

Automation builders

Connect change events to:

• Slack;
• Microsoft Teams;
• email workflows;
• n8n;
• Make;
• Google Sheets;
• databases;
• SEO dashboards;
• custom applications;
• AI analysis downstream.

---

How it works

For each term × source unit, the Actor:

1. validates and normalizes the seed term;
2. loads persistent state for that source-specific unit;
3. sends one autocomplete request;
4. validates the source response shape;
5. normalizes suggestions case-insensitively;
6. records each suggestion's 1-based position;
7. detects new suggestions;
8. updates consecutive-miss counters for absent suggestions;
9. confirms lost suggestions only after the configured miss threshold;
10. calculates rank deltas for suggestions present in both readings;
11. applies the event-type filter;
12. applies the result/runtime/billing caps;
13. writes admitted CHANGE_EVENT records;
14. advances state only after a successful source read;
15. writes a free RUN_SUMMARY;
16. stores operational STATS.

---

Input

Recommended example

{
  "terms": [
    "coffee maker",
    "crm software"
  ],
  "sources": [
    "google",
    "bing",
    "youtube"
  ],
  "language": "en",
  "country": "us",
  "eventTypes": [
    "NEW_KEYWORD",
    "LOST_KEYWORD",
    "RANK_CHANGE"
  ],
  "minRankDelta": 2,
  "lostAfterMisses": 2,
  "maxResults": 1000,
  "maxRuntimeMs": 300000,
  "requestDelayMs": 250
}

Input fields

Field	Default	Description
terms	required	Exact seed terms to monitor. Maximum 100 entries.
sources	["google"]	google, bing, and/or youtube.
language	en	Google/YouTube language such as en, pt, or pt-br.
country	us	Google/YouTube 2-letter country such as us, br, or de.
eventTypes	all three	Select which change events can be emitted.
minRankDelta	2	Minimum absolute position movement for RANK_CHANGE.
lostAfterMisses	2	Consecutive missing readings required before LOST_KEYWORD.
maxResults	1000	Maximum billable CHANGE_EVENT records.
maxRuntimeMs	300000	Hard runtime cap in milliseconds.
requestDelayMs	250	Request pacing delay before jitter. Range: 100–2,000 ms.
debug	false	Additional diagnostic logging.

---

Terms

Input terms are normalized by:

trim
lowercase
collapse repeated whitespace

Example:

"  CRM   Software "

becomes:

crm software

Duplicate normalized terms are monitored only once.

Empty entries are ignored.

If no valid term remains, the input is rejected before billing begins.

---

Sources

Example:

{
  "sources": [
    "google",
    "youtube"
  ]
}

The same term becomes two independent units:

google × term
youtube × term

Each has separate persistent state.

---

Language

Examples:

{
  "language": "en"
}

{
  "language": "pt"
}

{
  "language": "pt-br"
}

This setting affects Google and YouTube.

---

Country

Use exactly two letters:

{
  "country": "br"
}

Other examples:

us
gb
de
fr
ca
au

This setting affects Google and YouTube.

Bing's public route used here does not consume the country setting.

---

Event types

The Actor supports:

NEW_KEYWORD
LOST_KEYWORD
RANK_CHANGE

You can monitor all three or only the event types relevant to your workflow.

---

NEW_KEYWORD

A suggestion exists in the current reading and was not present in the previous successful state.

It is emitted immediately.

Example:

{
  "recordType": "CHANGE_EVENT",
  "eventType": "NEW_KEYWORD",
  "changeKind": "ADDED",
  "entityId": "kw:google/en-us/crm software/crm software ai",
  "source": "google",
  "observedAt": "2026-08-26T17:30:00.000Z",
  "previousObservedAt": "2026-08-25T17:30:00.000Z",
  "previous": null,
  "current": {
    "suggestion": "crm software ai",
    "position": 6
  },
  "term": "crm software",
  "monitorSource": "google"
}

---

Why new suggestions are immediate

A newly appearing suggestion is useful even if it later disappears.

The product therefore does not require two observations before emitting NEW_KEYWORD.

Losses use a stricter confirmation rule because a temporary missing suggestion can be source-side sampling noise.

---

LOST_KEYWORD

A suggestion is not declared lost after one missing reading by default.

Instead, its state keeps:

missesInARow

At the default:

{
  "lostAfterMisses": 2
}

the flow is:

run 1: suggestion present
run 2: missing → miss counter = 1 → no LOST_KEYWORD
run 3: missing → miss counter = 2 → LOST_KEYWORD

Example

{
  "recordType": "CHANGE_EVENT",
  "eventType": "LOST_KEYWORD",
  "changeKind": "REMOVED",
  "source": "google",
  "term": "crm software",
  "previous": {
    "suggestion": "crm software free",
    "position": 8
  },
  "current": null,
  "confirmedAfterMisses": 2,
  "lostAfterMissesApplied": 2
}

After a confirmed loss, the suggestion leaves persistent state.

If it later appears again, it becomes a legitimate NEW_KEYWORD.

---

Why loss confirmation exists

Autocomplete is not a transactional database.

A suggestion can temporarily disappear because of:

• source-side sampling;
• ranking reshuffling;
• transient source behavior;
• localization behavior.

Announcing a false loss can be more misleading than waiting for one additional reading.

The default anti-noise rule is therefore:

2 consecutive absences

You can configure:

1–10

according to your tolerance for latency versus noise.

---

RANK_CHANGE

A suggestion present in both readings can emit a rank event when:

absolute position delta >= minRankDelta

At the default:

{
  "minRankDelta": 2
}

these examples behave as follows:

position 4 → 5
delta = 1
ignored

position 8 → 2
delta = -6
RANK_CHANGE

Example

{
  "recordType": "CHANGE_EVENT",
  "eventType": "RANK_CHANGE",
  "changeKind": "CHANGED",
  "entityId": "kw:google/en-us/coffee maker/coffee maker walmart",
  "source": "google",
  "observedAt": "2026-08-16T21:14:09.812Z",
  "previousObservedAt": "2026-08-16T20:58:44.107Z",
  "changedFields": [
    "position"
  ],
  "previous": {
    "suggestion": "coffee maker walmart",
    "position": 4
  },
  "current": {
    "suggestion": "coffee maker walmart",
    "position": 1,
    "delta": -3
  },
  "term": "coffee maker",
  "monitorSource": "google",
  "minRankDeltaApplied": 2
}

---

Rank delta sign

For:

delta = current position - previous position

a negative delta means the suggestion moved closer to position 1.

Example:

8 → 2
delta = -6

A positive delta means it moved lower in the returned list.

---

Output

The default dataset contains:

CHANGE_EVENT
RUN_SUMMARY

CHANGE_EVENT is the billable monitoring event.

RUN_SUMMARY is free.

---

Core CHANGE_EVENT fields

Field	Description
recordType	CHANGE_EVENT.
eventType	NEW_KEYWORD, LOST_KEYWORD, or RANK_CHANGE.
changeKind	ADDED, REMOVED, or CHANGED.
entityId	Stable source/term/suggestion identity.
source	google, bing, or youtube.
observedAt	Current observation timestamp.
previousObservedAt	Previous successful observation timestamp.
changedFields	Changed canonical fields when applicable.
previous	Previous suggestion state.
current	Current suggestion state.
term	Normalized monitored term.
monitorSource	Source-specific monitored unit.
minRankDeltaApplied	Rank threshold applied to a RANK_CHANGE.
confirmedAfterMisses	Consecutive misses that confirmed a loss.
lostAfterMissesApplied	Loss-confirmation threshold.

---

RUN_SUMMARY

Every run attempts to write one free summary.

It can include:

recordsWritten
unitsRequested
unitsOk
unitsFailed
capReason
qualityAlert
sourceUnavailable
warnings
units
outcomeKind
eventsByType
baselinesCreated
requestsBySource
thresholds
locale
cost
pricingLabel

This keeps quiet monitoring runs visible even when no billable event occurs.

---

Per-unit summary data

The units array in RUN_SUMMARY can expose information such as:

source:term
status
suggestions
firstSight
newKeywords
lostKeywords
rankChanges
pendingMisses

This lets you distinguish:

no change

from:

source failure

or:

baseline creation

---

Request accounting

The design is intentionally lightweight:

1 source request
per
term × source unit

Retries can increase the actual HTTP request count when the source returns a retryable failure.

The free summary exposes:

requestsBySource

so request behavior remains auditable.

---

Pacing and retry behavior

The Actor uses controlled request pacing with jitter.

Default:

250 ms

Configurable range:

100–2000 ms

Retryable conditions can include:

• timeout;
• network failure;
• applicable HTTP 429/5xx conditions.

Unexpected successful-response shapes become a controlled:

API_CONTRACT_CHANGED

error rather than silently producing fake suggestions.

---

Redirects and source walls

The suggestion client does not automatically follow redirects on these API routes.

A redirect that behaves like a consent/login/access wall is treated as a blocked source condition.

The Actor does not rotate identities or attempt to bypass an access control.

---

Character encoding

Google suggestion responses can declare a non-UTF-8 character encoding for some locales.

The client reads the response charset from:

Content-Type

and decodes accordingly.

If the declared charset is unavailable, it falls back to UTF-8.

This helps prevent mojibake in multilingual suggestion monitoring.

---

Scheduling

This Actor is designed for recurring monitoring.

A practical starting point:

daily

The best cadence depends on how quickly you care about autocomplete changes.

More frequent runs can surface changes sooner, but autocomplete can also contain short-lived movement.

The anti-noise settings help control this.

---

Recommended scheduled workflow

1. Add the exact seed terms.
2. Select one or more sources.
3. Run once to establish free baselines.
4. Save the input as an Apify Task.
5. Open Schedules.
6. Choose the cadence.
7. Send future CHANGE_EVENT records downstream.

Possible destinations:

• Slack;
• Microsoft Teams;
• email workflows;
• Google Sheets;
• n8n;
• Make;
• webhooks;
• databases;
• SEO dashboards;
• internal applications.

---

Example SEO-monitoring input

{
  "terms": [
    "crm software",
    "project management software",
    "ai meeting notes"
  ],
  "sources": [
    "google"
  ],
  "language": "en",
  "country": "us",
  "minRankDelta": 2,
  "lostAfterMisses": 2,
  "maxResults": 200
}

A downstream workflow can:

• alert when a strategic autocomplete phrase appears;
• record a confirmed lost phrase;
• highlight large rank moves;
• maintain a change-history table;
• route selected events to an analyst;
• generate a weekly narrative with AI downstream.

---

Example YouTube topic monitoring

{
  "terms": [
    "ai tools",
    "home workout",
    "travel vlog"
  ],
  "sources": [
    "youtube"
  ],
  "language": "en",
  "country": "us",
  "eventTypes": [
    "NEW_KEYWORD",
    "RANK_CHANGE"
  ],
  "minRankDelta": 3
}

---

Example cross-source monitoring

{
  "terms": [
    "protein powder"
  ],
  "sources": [
    "google",
    "bing",
    "youtube"
  ]
}

This does not compare engines against each other.

It creates three independent timelines:

protein powder / google
protein powder / bing
protein powder / youtube

---

API

Run through the Apify API:

curl -s "https://api.apify.com/v2/acts/<YOUR_USERNAME>~keyword-trend-monitor/run-sync-get-dataset-items?token=<YOUR_APIFY_TOKEN>" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{
    "terms":["crm software","coffee maker"],
    "sources":["google","youtube"],
    "language":"en",
    "country":"us",
    "maxResults":200
  }'

Replace:

<YOUR_USERNAME>
<YOUR_APIFY_TOKEN>

with your Apify account values.

---

Integrations

Use with:

• Apify API;
• Tasks;
• Schedules;
• webhooks;
• n8n;
• Make;
• Google Sheets;
• Slack;
• Teams;
• databases;
• SEO dashboards;
• custom applications;
• AI agents downstream.

---

Pricing

This Actor uses Pay Per Event.

The code uses two custom billing events:

actor-start
trend-event

actor-start

Called once after valid input has been accepted.

Invalid input is rejected before the billing gate opens.

trend-event

Charged per delivered CHANGE_EVENT.

A billable event can be:

NEW_KEYWORD
LOST_KEYWORD
RANK_CHANGE

The free RUN_SUMMARY does not pass through event billing.

The Pricing tab on the Actor page is always the authoritative source for current prices.

---

What is free

Free by design:

first baseline
unchanged suggestions
unconfirmed first loss miss
rank movement below minRankDelta
RUN_SUMMARY

If a scheduled run has no qualifying changes, there is no trend-event result charge.

The configured start event can still apply.

---

Cost control

Primary controls:

terms
sources
eventTypes
minRankDelta
lostAfterMisses
maxResults
maxRuntimeMs
requestDelayMs

Fewer events

Use:

• fewer seed terms;
• fewer sources;
• a larger minRankDelta;
• only the event types you need.

Faster loss detection

Use:

{
  "lostAfterMisses": 1
}

but understand that this increases sensitivity to one-off source sampling.

Lower-noise loss detection

Use:

{
  "lostAfterMisses": 3
}

or higher.

The tradeoff is slower loss confirmation.

---

Run health

Operational statistics are persisted in:

STATS

in the default Key-Value Store.

Depending on the run, they can include:

• HTTP requests;
• retries;
• charged results;
• charge errors;
• units completed;
• units failed;
• quality alerts;
• source availability;
• runtime;
• caps;
• cost metrics.

The Actor also keeps historical health signals for source-volume anomalies.

---

Honest limits

Autocomplete is not a search-volume feed

The returned list is an autocomplete interface result.

Do not interpret position as monthly volume.

---

Rank is source order, not SEO ranking

position means position in the returned autocomplete suggestion list.

It is not:

• organic SERP position;
• YouTube video rank;
• Bing result rank;
• ad position.

---

Bing does not use the locale input in this implementation

The public Bing suggestion route used here does not receive language or country.

Google and YouTube do.

---

Source personalization and experimentation can exist

Public autocomplete behavior can vary because of source-side changes, experimentation, geography, timing, or other factors outside this Actor's control.

The Actor reports what the configured public route returned.

---

LOST_KEYWORD has intentional latency

At the default:

lostAfterMisses = 2

a loss requires two consecutive missing readings.

With a daily schedule, that can mean roughly one additional day before confirmation compared with a one-miss policy.

---

A source failure does not update state

Persistent state advances only after a successful read.

A failed request therefore does not become a false missing-suggestion observation.

---

Exact seed only

The Actor does not expand a seed through alphabet suffixes.

Use the companion keyword suggestion extractor when you need broad discovery.

---

The current Actor does not store a full historical rank series

Persistent state stores the latest suggestion positions and miss counters needed for diffing.

The Dataset is where emitted change history should be retained.

---

Autocomplete result count is source-controlled

The Actor does not promise a fixed number of suggestions per term.

The source decides how many suggestions it returns.

---

Public routes can change

Google, Bing, or YouTube can change:

• response shape;
• rate limits;
• redirect behavior;
• encoding;
• suggestion logic.

The Actor includes explicit contract checks and controlled errors to surface such changes.

---

No sentiment or intent classification

The Actor does not classify autocomplete suggestions as:

• positive;
• negative;
• commercial;
• informational;
• transactional.

Use a separate downstream analysis step when needed.

---

No AI in runtime

All diff decisions are deterministic.

No LLM creates or interprets suggestions.

---

FAQ

Do I need a Google account?

No.

Do I need a Bing account?

No.

Do I need a YouTube account?

No.

Do I need API keys?

No for the public suggestion routes used by this Actor.

Does it use browser automation?

No.

What happens on the first run?

Each term × source creates a baseline.

No baseline suggestion is billed as a change event.

Can I monitor 100 terms?

Yes.

The input accepts up to 100 normalized terms.

Can I monitor all three sources?

Yes.

Are Google, Bing and YouTube compared with each other?

No.

They are independent state timelines.

Can I use Brazilian Portuguese?

Yes.

For Google/YouTube:

{
  "language": "pt-br",
  "country": "br"
}

Does the country affect Bing?

No.

What is a NEW_KEYWORD?

A suggestion present now that was not present in the previous successful state.

What is a LOST_KEYWORD?

A previously tracked suggestion that remained absent for the configured number of consecutive successful readings.

Why did a suggestion disappear but no loss event fire?

At the default, the first absence only increments the miss counter.

A second consecutive absence confirms the loss.

What is RANK_CHANGE?

A suggestion was present in both readings and moved at least minRankDelta positions.

Is rank change the same as Google Trends?

No.

Does the Actor return search volume?

No.

Does it return CPC?

No.

Does it return keyword difficulty?

No.

Can I schedule it?

Yes.

Daily is a practical starting cadence.

Are quiet runs charged per result?

No trend-event is created when there is no qualifying change.

The configured run-start event can still apply.

Is the summary charged?

No.

What am I charged for?

The custom start event shown in Pricing plus delivered CHANGE_EVENT records.

Is this affiliated with Google, Microsoft, Bing or YouTube?

No.

This is an independent community Actor using public autocomplete/suggestion routes.

---

Support

For bugs, questions, or requested fields:

johnatan291303@gmail.com

You can also use the Issues tab on the Actor page.

---

Part of the JM Forge suite

Also from the same developer:

• Google, Bing & YouTube Keyword Suggest Scraper — broad keyword discovery and autocomplete expansion.
• Brand News Monitor & GDELT Mention Alerts — stateful global brand-news monitoring.
• ATS Salary, Skills & Seniority Intelligence — structured hiring intelligence from public ATS boards.

JM Forge Actors remain independent tools.

Use the extractor for broad keyword discovery.

Use this monitor for recurring autocomplete change intelligence.