Google Maps Lead Lists by Zip Code and Category, Written Straight to Sheets

William's n8n Google Maps lead build writes business lists to Sheets by zip code by reading unfinished zip rows and category rows from Google Sheets, POSTing each pair as a Places Text Search (New) textQuery, then append-or-updating a Results tab on place_id. The canvas in my library is named Google Maps Leads | Sheets -> Maps -> Sheets (workflow id h1mijH34S11fmgnI, created 15 March 2025, last saved 22 May 2025). It is a list builder. It is not a CRM, not an outreach sender, and not a Google Business Profile job.

I am William Spurlock, founder, AI Systems Architect, and Fractional AI CTO. I have shipped 600+ automations with 500+ live, logged 20,000+ hours inside agentic systems, and deleted 35,000+ hours of client busywork across that book of work. I built this stack. I am not going to invent a client name, a weekly lead count, or a dollar return you cannot audit.

This spoke sits under the first AI automation every small business should build. That parent owns intake: form to record to confirmation. This page owns one later receipt: zip × category in, business rows out. If you still copy "plumbers 85001" by hand into a tab, this is the shape I use instead.

There is no model node on this canvas. No Claude Opus 4.8. No GPT-5.5. No Gemini 3.1 Pro. Rules read Sheets. An HTTP node hits Places. Rules write Sheets. If you want a model to draft a first-touch email later, that is a different workflow and a different approval gate.

I keep two other Maps canvases in the same library. I am not writing them here. One chats. One calls Apify. This page is only Sheets → Places → Sheets. If your operator opens the wrong workflow, they will look for a chat trigger that does not exist on this canvas.

Credentials the owner has to keep alive #

Credential	Used by	What dies when it expires
Google Sheets OAuth (or service account)	Get Zip Codes, Get Subcategory, Add rows, Update Status, GS - Get Status	Reads stop. Writes stop. Backoff cannot save a revoked token.
HTTP header auth for Places	GMaps API	Every textQuery 401s / 403s. Inner loop still ticks. Results stay thin.

Rotate those in n8n credentials. Do not paste keys into Settings. Settings is sheet names, not secrets.

---

How does William's n8n Google Maps lead build write business lists to Sheets by zip code? #

It nested-loops unfinished zips against allowed categories, calls places:searchText with "{Subcategory} {zip}", unnests the places array, drops duplicate place ids inside that response, and writes the Results tab with Google Sheets appendOrUpdate matched on place_id. Then it stamps the zip row scraped and asks the inner loop for the next category.

The Airtable library card still says "scrape." The live node does not open Maps HTML. It POSTs to https://places.googleapis.com/v1/places:searchText with header auth and a required X-Goog-FieldMask. That is the official Places API (New) Text Search method Google documents as of the usage and billing page last updated 10 September 2026.

Three ways this canvas starts #

I left three triggers on the same Settings node. You pick one. You do not fire all three at once.

Node	Type	What it does on this canvas
When clicking "Execute Workflow"	Manual trigger	Owner-run. This is where I start a new zip list.
Execute Workflow Trigger	Sub-workflow start	Another canvas can call this one with no chat UI.
Run workflow every hours	Schedule trigger	Interval in the saved JSON is 15 minutes. That is a convenience, not a quota plan.

I would rather run this from the manual trigger with the Limit node still at 3 zips until the Results tab looks honest. The 15-minute schedule is how you burn Places SKUs and Sheets write quota while you are still spelling category names wrong.

The nested loop, node by node #

The outer loop is zip. The inner loop is category. Settings holds three strings so I am not editing five Google Sheets nodes when the workbook moves.

Node	Job	Owner-facing fact
Settings	Set gs_url, sheet, catSheet	Zip tab name in this canvas is AZ Zips. Category tab is Google Maps Categories. I am not publishing the live spreadsheet URL.
GS - Get Zip Codes	Read the zip tab	Reads every row. Filtering happens next.
Zips	Coerce zip to a number	Bad types here become quiet misses later.
Filter Zips	Keep rows whose status is empty	A filled status cell is a skip. This is the resume switch.
Set Row Number → Split Out → Limit	Prep the outer loop	Limit is saved at maxItems: 3. Forget that and you will swear the list is broken.
Loop Zips	SplitInBatches on zips	One zip in flight.
GS - Get Subcategory	Read the category tab	Runs inside the zip loop. Same category list for every zip.
Filter Subcategories	Keep rows where STATUS is not Ignore	Park a vertical without deleting the row.
Subcategory	Set the current Subcategory string	This string is half of the Places query.
Loop Subcats	SplitInBatches on categories	One category in flight for the current zip.
Set Zip	Pull zip from Loop Zips	Inner items do not inherit the outer zip unless you set it.

After Set Zip, GMaps API builds one query. The body field is textQuery. The expression in the saved JSON is the current subcategory, a space, then the current zip. Google's own Text Search examples look like "pizza in New York." This canvas uses the shorter "Plumbers 85001" shape. Google says identical queries are not guaranteed to return the same list. Treat the Results tab as a snapshot of that call, not a census of the zip.

What the Places node asks for #

The HTTP node is named GMaps API. Method POST. Full response on. Auth is a generic HTTP header credential — the API key stays in n8n credentials, not in the sheet.

Request piece	Value on this canvas	Why it matters
URL	https://places.googleapis.com/v1/places:searchText	Places API (New) Text Search.
textQuery	Subcategory + space + zip	One pair, one call.
X-Goog-FieldMask	places.id, displayName, addressComponents, formattedAddress, primaryType, primaryTypeDisplayName, types, location, nationalPhoneNumber, rating, userRatingCount, websiteUri, editorialSummary, reviews, attributions	No FieldMask means an error. You are billed at the highest SKU in the mask.
Pagination	None	One POST per pair. No pageToken loop. You get one page, not every business Google could list.

I left reviews and editorialSummary on that mask. That is an owner choice, not a free extra. Google bills Text Search at Essentials, Pro, or Enterprise (including Enterprise + Atmosphere) based on the fields you ask for. Read the current Places SKU table before you copy this mask onto a long zip list. If you only need name, phone, website, and address, trim the mask. I am not going to price your project from a blog post.

Empty body, unnest, write #

If Empty checks whether $json.body is an empty object. Empty → back to Loop Subcats. No Results row. No drama. That is how a dead combo exits.

If the body has places, Place Array (Code node) walks items[0].json.body.places and emits one n8n item per place. Set Place ID copies place.id onto places.id. Remove Duplicates compares place.id and places.id so the same place does not write twice inside that response.

Add rows in Google Sheets is appendOrUpdate on the tab named Results. The match column is place_id. Same place found under a second category updates the row. It does not invent a second lead.

On success, the write path hits GS - Get Status, then Update Status to Success. That update writes status: scraped onto the zip tab, matching zip only, and also writes the last subcat it finished. Then it returns to Loop Subcats.

That last sentence is the resume trap. Status is zip-level. The inner loop can still finish remaining categories in the same run. The next run will skip that zip as soon as status is filled. If the run dies after the first category succeeds, leftover categories for that zip stay uncalled until you clear the status cell.

---

What sits in the zip and category sheets before a run? #

Two tabs in one workbook, plus a Results tab the write node already expects. Settings points every Sheets node at the same document URL. You change tab names in Settings, not by hunting the canvas.

I am not dropping the production spreadsheet link from this receipt. You will use your own workbook. The tab names below are the ones saved in this JSON.

Zip tab (AZ Zips in this canvas) #

The name is Arizona-specific because that book of work started as an Arizona list. Rename the tab and the Settings sheet value if your list is Ohio. The filter does not care what the tab is called. It cares about status.

Column	Who writes it	What this canvas does with it
zip	You, before the run	Outer loop key. Coerced to a number. Used in textQuery and in the status update match.
status	You leave it blank; the workflow writes scraped	Filter Zips keeps empty cells only.
subcat	Workflow, on success	Last subcategory that completed for that zip. Overwritten each inner success.
row_number	Sheets / n8n	Tracking. Treat it as read-only.

Owner rules I actually use:

1. One zip per row. No "85001-85004" ranges in a single cell.
2. Leave status empty for work you want done. Type anything else and that zip is skipped.
3. Do not pre-fill subcat. That column is a breadcrumb, not a queue.
4. If you need to redo a zip, clear status. Do not delete the Results rows unless you also want those place_id keys gone.

Category tab (Google Maps Categories) #

Column	Who writes it	What this canvas does with it
Subcategory	You	Exact string in the Places textQuery. Spelling is the query.
STATUS	You	Row is used when this is not Ignore.

Ignore is how I park "roofing" for a week without deleting the vertical. Any other value, including blank, passes the filter.

Write categories the way you would type them into Maps: Plumbers, HVAC contractors, Landscapers. Do not stuff three verticals in one cell. The inner loop is one string per call.

Results tab (Results) #

Create the tab before the first write. The append-or-update node maps columns by name. If the header is missing, the write fails and the backoff path starts.

You do not type businesses into Results by hand. You type zips and categories. Results is the output.

What you do not put in these sheets #

Temptation	Why it does not belong here
Decision-maker first name	This canvas never writes firstname / lastname. Those headers can exist on Results and stay empty.
Email	Same. The Places mapping does not include email.
"Do not contact" flags	Add them later in your CRM. This workflow will overwrite a Results row on place_id if the place comes back again.
Outreach copy	Wrong layer. This is a list.

If the next hop is HubSpot or Airtable, wire that in a second workflow after a human has looked at Results. The hour-or-less connect pattern is in how to connect n8n to your CRM, email, and website.

Settings I actually edit #

The Settings node is a Set node, not a second spreadsheet. Three assignments, then every Sheets node reads them.

Assignment	Saved value on this canvas	What you change
gs_url	A Google Sheets document URL	Your workbook. I am not reprinting the production link.
sheet	AZ Zips	Zip tab name. Must match the tab exactly.
catSheet	Google Maps Categories	Category tab name. Same rule.

Results is not in Settings. The write node hardcodes the tab name Results. If you rename that tab, edit Add rows in Google Sheets, not Settings.

A first run that cannot wreck the list #

This is the sequence I use on a new workbook. It is slow on purpose.

1. Put two zips in AZ Zips (or your renamed tab). Leave status empty.
2. Put two categories on the category tab. Set a third row to Ignore so you can prove the filter.
3. Leave Limit at 3. You only have two zips anyway.
4. Fire the manual trigger. Do not enable the 15-minute schedule.
5. Watch GMaps API once. Confirm textQuery is Category Zip and not an empty string.
6. Open Results. You should see place_id and title on any combo that returned places.
7. Confirm the first zip that finished a category now says scraped.
8. Clear that status if you still need the second category and the run died early.

If step 5 shows a 403 or a missing FieldMask error, stop. That is a Cloud project problem, not an n8n mapping problem.

Manual vs this canvas #

Step	By hand	This canvas
Pick a zip	You remember which ones you finished	Empty status is the queue
Pick a category	You retype it	Inner loop reads the category tab
Search	You type into Maps	One textQuery POST
Copy name, phone, site	You paste	Write node maps ten fields
Skip a repeat	You eyeball the sheet	place_id append-or-update
Mark the zip done	You highlight a row	status = scraped
Survives a Sheet quota blip	You wait and try again	Backoff 1 / 2 / 4 / 8 / 16 seconds, five tries

Hand work still wins when you have five businesses in one town and you already know them. This canvas wins when the zip list is longer than your patience and you are willing to accept one Text Search page per pair.

---

Which columns land on the Results tab? #

The write node fills place_id, title, phone, website, rating, reviews, address, type, gps_coordinates, and types. Everything else on that tab is a header I left for later hops. Empty is not a bug.

Columns this canvas actually maps #

Results column	Expression in the write node	Source
place_id	$json.place.id	Places id. Match key.
title	$json.place.displayName.text	Display name.
phone	$json.place.nationalPhoneNumber	National-format phone when Google returns one.
website	$json.place.websiteUri	Website URI when Google returns one.
rating	$json.place.rating	Rating. Higher SKU field.
reviews	$json.place.reviews	Review objects when returned. Highest-cost field on this mask.
address	$json.place.formattedAddress	Formatted address.
type	Current Subcategory	Your category string, not Google's primaryType.
gps_coordinates	JSON with latitude / longitude	Places location.
types	$json.place.types	Google type array.

type is the search vertical you asked for. types is what Places attached to the place. They will disagree. Keep both. When I am hunting HVAC contractors and Google tags a place as a plumber, I want to see that conflict in the sheet, not hide it.

Phone, site, and rating are optional on Google's side #

Places does not owe you a complete card. I have run combos where title and address land and phone or websiteUri stay empty. The write node still upserts the row. That is a useful row. It is not a complete outreach row.

Field	When it is often empty	What I do
phone	Service-area businesses, or a listing Google never collected a number for	Leave it. Do not invent a number from a website later without a second, approved hop.
website	Places with no URI	Same.
rating	New or unrated listings	Keep the row. Rating is not a quality gate on this canvas.
reviews	Mask asked; Google returned none	Empty cell. You still paid for the SKU the mask requested.

If your seller needs phone-or-skip, filter Results after the run. Do not add that filter inside the Places node. You will hide listings you might still want as a market map.

Headers that exist and stay blank #

The Results schema on this node also lists ACTION, STATUS, email, name, firstname, lastname, clean url, WP API, WP, facebook, and instagram. This workflow does not fill them. The Upwork-style library blurb that mentions email is marketing copy. The mapping does not.

If a later enrichment canvas writes email into that column, fine. Do not tell a seller this Text Search call harvested inboxes. It did not.

Dedup and update behavior #

Event	What happens
New place_id	Append a Results row.
Same place_id again	Update the matched row. Phone, site, rating, address refresh.
Same business, new Places id	Second row. I have seen this when Google recatalogs a listing. You merge by hand.
Two categories hit the same place in one zip	One row. type becomes whichever category wrote last.

That last row is why I do not treat type as gospel. If you need a zip × category membership table, you need a second key. This canvas is a place list, not a combo list.

What one page of Text Search is not #

Text Search returns a places array. This canvas does not follow pageToken. It does not set maxResultCount in the saved body. You get whatever that single POST returned.

I will not dress that up as "every business in the zip." It is every place Google put on that page for that query, on that run. If you need a fuller pull, you are in product-and-terms territory with Google, not in a blog comment asking me for a paging recipe.

---

What fails first, and what Google will not let you do? #

Empty Places bodies, Sheets write quota, a scraped stamp that is too eager, and a FieldMask that bills Enterprise while you still think this is a cheap id-only search. The canvas already retries Sheets writes. It does not invent a way around Google's quota or terms.

Failure modes I watch as the owner #

Failure	What you see	What I do
Limit still at 3	Three zips processed. The rest wait with empty status.	Raise maxItems only after a clean test.
Category STATUS = Ignore	That vertical never calls Places.	Expected. Not an error.
Zip status already filled	Zip skipped at Filter Zips.	Clear the cell to redo.
Places body empty	If Empty returns to Loop Subcats.	Leave it. The combo had nothing to write.
HTTP 4xx / 5xx on Places	The HTTP node errors. No Results write for that pair.	Check credential, FieldMask, and billing on the Google Cloud project. Do not "retry harder" in a tight loop.
Sheets WriteGroup quota	Add rows / Update Status take the error output.	Backoff path. See below.
Run dies mid-zip	Zip may already be scraped after the first successful category.	Clear status if leftover categories still matter.
Duplicate place, new id	Two Results rows.	Manual merge.
Schedule left at 15 minutes on a long list	Places QPM and Sheets writes stack.	Turn the schedule off until the list is small and the mask is trimmed.

The backoff that is already on the canvas #

Three write-adjacent paths use the same idea. Exponential Backoff (Code) reads retryCount or starts at 0. maxRetries is 5. initialDelay is 1 second. Wait time is 1 * 2^retryCount seconds: 1, then 2, then 4, then 8, then 16. Wait sleeps that many seconds. Check Max Retries either loops back to the Sheets node or hits Stop and Error.

That path exists because Google Sheets will tell you the write group is exhausted. It is a polite pause. It is not a method for beating Places QPM, and I am not going to publish one.

Places API (New) rate limits are per method per project. Text Search has its own bucket. Sheets has its own. Watch both. When a method hits quota, the service stops answering. Raising a Cloud quota is a console decision you make after you have read the bill. I will not walk that click-path here.

Terms I treat as owner constraints #

I am not your counsel. I am the person who has to keep this canvas from becoming a Terms problem.

Google's Maps Platform Service Specific Terms (Places section, as published for non-EEA billing accounts; last modified 10 June 2026 on that page) say, in short:

Rule	What it means on this canvas
Use the official API	This HTTP node is Places Text Search. Do not swap it for an unofficial Maps HTML pull.
FieldMask required	Already on the node. Do not delete it.
Place IDs may be stored	place_id as the Results match key is the intended store.
Lat/lng cache cap	Places latitude and longitude may be cached up to 30 consecutive calendar days, then deleted. gps_coordinates is not a forever column.
Attribution	You attribute Maps content per Google's documentation. A private ops sheet still has an owner who must follow the docs.
No Places content on a non-Google map	Do not plot these rows on a rival map product.
Highest SKU wins	reviews on the mask is how a "cheap list" becomes an Enterprise line item.
Monthly credit	The $200 Maps credit Google described on the billing page applied through 28 February 2025. Do not budget 2026 runs on a credit that already ended.

Google also points from that billing page to Places policies and the License Restrictions in the Google Maps Platform Terms of Service. Read those before you schedule anything. If your billing address is in the EEA, Google publishes a separate EEA terms set. I am not going to paraphrase that into a loophole.

What I will not do in this post: give you a browser scrape, a header spoof, a proxy pool, or a "just page until the zip is empty" recipe. If the official API and your quota cannot support the list you want, the list is too big for this canvas.

Node inventory (this canvas only) #

I counted the saved JSON. Sticky notes and the three triggers sit beside the work nodes below.

Family	Nodes on this canvas
Triggers	When clicking "Execute Workflow"; Execute Workflow Trigger; Run workflow every hours
Config	Settings
Sheets read	GS - Get Zip Codes; GS - Get Subcategory; GS - Get Status
Sheets write	Add rows in Google Sheets; Update Status to Success
Loop / filter	Filter Zips; Filter Subcategories; Set Row Number; Split Out; Limit; Loop Zips; Loop Subcats; Zips; Subcategory; Set Zip
Places	GMaps API; If Empty; Place Array; Set Place ID; Remove Duplicates
Backoff	Exponential Backoff; Exponential Backoff1; Exponential Backoff2; Wait; Wait1; Wait2; Check Max Retries; Check Max Retries1; Check Max Retries2; Stop and Error; Stop and Error1; Stop and Error2

If a rebuild is missing Limit, Filter Zips, or the place_id match, it is not this receipt. It is a new canvas that will re-hit finished zips or duplicate Results rows.

When I refuse to turn this on #

I do not activate the schedule when any of these are true:

• The FieldMask still asks for reviews and nobody has looked at this month's Places SKU table.
• The zip tab has hundreds of empty-status rows and Limit is already raised.
• The operator wants emails, decision-maker names, or a "complete zip" guarantee from one POST.
• Someone asked to point the HTTP node at a non-Google Maps HTML source.
• Results is the same tab the sales team already uses as a live dialer with no copy.

Those are not moral lectures. They are how this canvas stays a list builder instead of a Terms incident.

What this receipt is not #

Not this	Why
Google Business Profile work	Different product. Different terms. This page does not own local AI visibility.
A CRM	Results is a sheet. Contacts, stages, and do-not-contact live elsewhere.
An email finder	Email is an empty header on this write.
A Zapier clone of the same nested loop	You can start in Zapier. You will feel the zip × category fan-out. I said my piece in n8n vs Make vs Zapier in 2026.
A model that scores "hot leads"	No LLM node. If you add one later, keep it off send.

---

FAQ: Zip lists, Places fields, and resume #

Does this scrape Google Maps HTML? #

No. The live node is an official Places Text Search POST to places.googleapis.com. The library brief still says "scrape" as shorthand. HTML collection is a Terms problem I will not help with. Use the API, a FieldMask, and a billed Cloud project.

Why is the zip tab named AZ Zips? #

Because this canvas started as an Arizona zip list, and Settings hardcodes the tab name AZ Zips. Change Settings sheet when you rename the tab. Filter Zips only cares that status is empty. The letters AZ are not a geo-filter inside Places. The zip in textQuery is.

What happens when Places returns nothing for a zip and category? #

If Empty sends the item back to Loop Subcats. No Results row is written. The zip is not stamped scraped on that empty path. A later category for the same zip can still write. If every category comes back empty and you never hit Update Status, the zip stays eligible next run.

Does the Results sheet get emails from this Places workflow? #

No. The append-or-update mapping does not write email. The Results tab can show an email header from an older schema. This canvas leaves it blank. Anyone who tells you Text Search filled inboxes from this node is reading the sales blurb, not the write node.

How do you resume a half-finished zip list? #

Filter Zips only lets empty status through, and Update Status stamps the whole zip scraped after a successful category write. Same-run inner categories still finish. A crash after the first success skips leftover categories on the next run until you clear status. There is no zip × category status grain on this canvas.

Can I run this Places lead build every 15 minutes? #

The saved schedule node is 15 minutes. I do not leave that on for a long zip list. Places QPM is per method per project. Sheets writes have their own quota. Backoff handles a burst of Sheet errors. It does not make a 15-minute hammer safe. Manual runs with Limit at 3 until the tab looks right.

Is this the same as importing the list into a CRM? #

No. This workflow stops at Google Sheets. place_id is a good upsert key if you later create contacts. Stages, owners, and suppression lists do not belong in Results. Connect CRM in a second workflow after a human has scanned the tab.

What is place_id used for here? #

It is the append-or-update match and the Google ID this canvas is allowed to store. Google's service terms allow caching Place IDs. That is why I match on place_id instead of name plus phone. Names change. Ids are the join.

Do I need n8n Cloud, or can I self-host? #

Either. The nodes are stock: Sheets, HTTP Request, SplitInBatches, If, Code, Wait. Credentials still have to reach Google Sheets and Places. Self-host if you want the credential on your box. Cloud if you want Google to keep the process up. The business outcome is the same list.

---

Book an AI automation strategy call #

This canvas does one job well: zip and category in, Places page out, Results tab updated, zip stamped. It does not send mail. It does not score intent. It does not replace the intake path I still tell people to build first.

If your first automation is still "copy the form into a sheet by hand," go back to the first AI automation every small business should build and ship that. If the form already lands and you are the person wasting a morning on Maps copy-paste, this is the next receipt I actually run.

Keep the Limit small. Trim the FieldMask. Read Google's current Places SKU table and the 30-day lat/lng rule. Clear status on purpose. Do not ask this workflow to become a shadow Maps product.

If you want a Sheets-to-Places-to-Sheets list builder mapped to your real zip tab, category list, and CRM hop — not a generic tool tour — book an AI automation strategy call. We will pick the first canvas, keep Places on the official API, and leave send off until the rows are ones you would dial yourself.