NLP SEO tools: test your draft with Google's Natural Language API
Short answer Google's Cloud Natural Language API shows which people, places and things a machine finds in your draft, which one it rates as central, and what category the text falls in. It's a quick clarity check, free for the first 5,000 units a month. It is not Google Search: no tool can show how Search scores your page.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhat does an NLP tool tell you about your draft?
It tells you what a language model thinks your text is about. Paste in a service page and Google's Cloud Natural Language API returns these results (Google Cloud: Natural Language API basics):
| Feature | What it returns | What to check on a draft |
|---|---|---|
| Entity analysis | The people, places, businesses, products and common nouns it finds, each with a type | Are your service, city and business name all found? |
| Salience | A 0 to 1 score for how central each entity is to the whole text | Is your main service near the top? |
| Content classification | Categories such as "/Home & Garden/..." from Google's fixed list | Does the category match what the page sells? |
| Sentiment | Score from -1.0 (negative) to 1.0 (positive), plus magnitude (how much emotion) | Useful for reviews; rarely matters for a service page |
| Syntax | Sentences, parts of speech, how words connect | Mostly for developers |
Google's reference defines salience this way:
"The salience score for an entity provides information about the importance or centrality of that entity to the entire document text." (Google Cloud: Entity reference (v1))
For most entities it also returns a Wikipedia link and a Knowledge Graph ID "if they are available" (same page). That shows the model matched a word to a known thing, such as "Phoenix" the city rather than the bird. For what the Knowledge Graph is, see knowledge graph.
Is this the NLP that Google Search uses?
No one outside Google can say, and Google doesn't claim it. The Natural Language API is a Google Cloud product sold to developers. Google Search's documentation doesn't say Search uses it, and warns about third-party tools that claim to use Google's internal metrics:
"No third-party tool has access to our internal ranking or AI systems." (Google Search Central: Optimizing your website for generative AI features on Google Search)
So read the API as a second reader, not a scorecard. If a machine can't tell your page is about drain cleaning in Tucson, a person skimming on a phone may not either. That's the use. A salience of 0.4 versus 0.6 is not a ranking number. How Google's own language systems read a page is covered in NLP for SEO.
How to run a draft through Google's Natural Language API
There are two ways in: the demo for a quick look, or your own Google Cloud project for full results.
Quick look. Google's product page has a "Natural Language API demo" section where you can try the API on your own text (Google Cloud: Natural Language AI). Paste one section at a time and read the entity list.
Full results, as of November 2026 (Google Cloud: Natural Language API quickstart and setup guide):
- Sign in with a Google account and create a Google Cloud project.
- Turn on billing for the project. The setup guide requires a billing account, even if you stay inside the free tier.
- Enable the Cloud Natural Language API for the project.
- Install the Google Cloud CLI (the command-line tool) and run `gcloud init`.
- Run the entity command from Google's quickstart, with your draft as the text: `gcloud ml language analyze-entities --content="your draft text"`.
- Read the result: each entity's `name`, `type`, `salience` and `metadata` (the Wikipedia link, when there is one).
One version detail matters. The v1 API returns `salience`. The newer v2 API's entity record has no salience field (Google Cloud: Entity reference (v2)). If you call the API directly, use the v1 address (`language.googleapis.com/v1/documents:analyzeEntities`) shown in the quickstart.
How to read the results
Check three things, in this order:
- The top entity. Sort by salience. On a page about roof repair in Denver, "roof repair" or "roof" should lead. If "company", "team" or "years" lead, the copy talks about you more than the job. Rewrite the opening to name the service.
- Missing entities. Your city, the service and your business name should each appear. If the city is missing, the page probably never says it in plain words. Add it where it's true: "We repair roofs in Denver and Aurora."
- The category. If a dental implant page comes back as a general health or beauty category, the page may never say "dental implant" plainly. Compare against Google's category list (Google Cloud: Content categories).
What not to do: repeat a word until its salience rises. Filling a page with keywords "in an attempt to manipulate rankings" is keyword stuffing under Google's spam policies (Google Search Central: Spam policies). Fix the meaning, then stop.
What does the Natural Language API cost?
For one owner checking a few pages, nothing. Prices from Google's pricing page, in US dollars (Google Cloud: Natural Language API pricing):
| Feature | Free each month | Then, per 1,000 units |
|---|---|---|
| Entity analysis | 5,000 units | $1.00 |
| Sentiment analysis | 5,000 units | $1.00 |
| Entity sentiment | 5,000 units | $2.00 |
| Syntax | 5,000 units | $0.50 |
| Content classification | 30,000 units | $2.00 |
A unit is 1,000 characters, rounded up per request, and spaces and HTML tags count. A 6,000-character page is 6 units, so the free tier covers hundreds of checks a month. New Google Cloud customers also get "up to $300 in free credits" (Google Cloud: Natural Language AI).
In Canada, Google says that if you pay in a currency other than US dollars, the prices listed in your currency on its SKU list apply (same pricing page).
Other NLP and entity tools, and the free option
You may not need the API at all. The choices:
| Option | What you get | Limit |
|---|---|---|
| Natural Language API | Google's entity, salience and category output | Needs a Cloud project with billing; developer-style output |
| Content editors with "NLP terms" | A list of words that top-ranking pages use | A word list, not meaning. Paid subscriptions |
| An AI chat assistant | Ask "What is this page about, and what's missing for a customer?" | Answers vary run to run; check its claims |
| Search Console (free) | The real searches your page appears for | Shows results after publishing, not before |
The free check that matters most is the last one. A few weeks after publishing, Search Console's Performance report lists the queries your page shows up for. If they match the page's topic, Google read it as you meant. See Google Search Console for a small business.
Example
Illustrative cases with invented numbers.
A landscaper's lawn care page (US dollars)
The API's top entity is "family", from an opening about a family-owned company since 1995. "Lawn care" comes fourth and the city doesn't appear. The owner rewrites the first two lines to "Weekly lawn mowing and fertilizing in Raleigh, from $45 a visit." On the next run, "lawn care" leads and "Raleigh" shows with its Wikipedia link. Cost: 3 units, inside the free tier.
An accountant in Calgary (Canadian dollars)
A page on small business bookkeeping comes back classified under a general finance category, and "GST" is missing because the page only says "sales tax". The accountant names "GST returns" and "T2 corporate returns" in the services list. If the account pays in Canadian dollars, Google's Canadian-dollar SKU prices apply above the free tier; this check stays free.
FAQ
Is salience a Google ranking factor?
Not that Google has said. Salience is a score the Cloud Natural Language API calculates for its own users. Google Search's documentation doesn't mention it.
Can I use the Natural Language API without coding?
For a quick look, yes: the demo on the product page. For full, repeatable output, you need the command-line tool or a short API call.
Will the API recognize my business name?
Usually, as an entity. A Wikipedia link appears only "if available", so most small businesses get none. That's normal and says nothing about rankings.
Are "NLP terms" from content tools worth adding?
Treat them as questions your page might skip, not words to insert. Add a term only if a customer expects the page to cover it.
Different for you?
- You want to know how Google itself reads language: NLP for SEO.
- You want Google to see your business as a known entity: entity SEO and your Google knowledge panel.
- The page reads clearly but still doesn't rank: SEO content optimization.
- You'd rather have pages written and checked for you: SEO & GEO Agent.