How to use ChatGPT for SEO keyword research

Find the keywords people are actually searching for, see which ones are rising, and read the pages that already rank, all from one ChatGPT conversation. Install the Firecrawl plugin and ChatGPT can do keyword research with Trends search-interest data and Similarweb traffic figures from Firecrawl Alexandria, then scrape the top results for a query.

• Content marketers building a topic cluster around one seed term.
• SEO leads deciding which of two phrasings to target, and when in the year to publish.
• Founders using ChatGPT for SEO research on a competitor before picking a niche.

Plugin: the Firecrawl plugin. Besides calling Firecrawl Alexandria providers, it runs web searches and scrapes result pages into clean text, so ChatGPT can read what ranks for a keyword. Each call is billed against your Firecrawl balance.

Explore Firecrawl Alexandria data providers across categories

Swap in your own seed keyword and country:

@firecrawl Help me do keyword research for "standing desk" in the US.
Use Trends in Firecrawl Alexandria and Firecrawl search, and show me the
price of each tool before running it.
 
Give me:
- The top 15 related queries and the rising ones from the past 12 months,
  with their scores.
- Interest over the past 12 months for my two best candidates, so I can
  see seasonality.
- The pages that come up for my top candidate, scraped, with each page's
  H1, H2s and approximate length.
 
Treat Trends scores as relative interest. Don't invent monthly search
volumes or difficulty scores.

• Get the plugin. Use the install button under these steps.
• Connect your Firecrawl account. Trends lookups and page scrapes use a handful of credits each. Create a free account if you don't have one.
• Mind the Similarweb allowance. Similarweb calls cost more than other tools and also count toward a separate monthly Similarweb limit for your team.

Keyword research is a numbers exercise, and ChatGPT keyword research without data runs into five problems.

1. Search volumes it can't see. Semrush points out that chatbots lack direct access to search engine data, so they can't reliably say how popular a keyword is or how hard it is to rank for (Semrush). Ahrefs found the same: a model will produce a convincing table of volumes and CPCs, and those figures are guesses (Ahrefs).
2. Demand it learned months ago. OpenAI says responses don't reflect events past the training cutoff when no tools are involved (OpenAI), so a query that started rising this quarter won't be on its list.
3. Seasonality goes unnoticed. Google says demand for queries can shift with new trends or across the seasons, and suggests Google Trends to check whether a drop is yours or the whole web's (Google Search Central). A static keyword list carries no timing.
4. No view of the results page. Semrush recommends reviewing the top-ranking results for a keyword before you target it (Semrush). ChatGPT's suggestions don't show you which pages rank or what format wins.
5. Search answers can be partial. Even with web search on, OpenAI says results and citations can be incomplete, outdated or incorrect (OpenAI).

Using ChatGPT for keyword research with Firecrawl works like this:

• Firecrawl Alexandria: one catalog with 100+ official providers paid under agreements, plus indexes Firecrawl builds over the open web, academic papers, docs for developers and government publications. Firecrawl ran 845 tasks with an unchanged model and prompts, and agents with Firecrawl Alexandria delivered answers 21% better than those relying on default web tools.
• Trends for demand: related and rising queries widen a seed term into a list, and interest over time shows when people search. For "standing desk" in the US over the past year, "electric standing desk" scored 100 and "adjustable standing desk" 87, while "electric adjustable standing desk" was up 550%.
• Search and scrape for the results page: Firecrawl searches the keyword and returns the content of the pages it finds, so ChatGPT can compare headings, length and format.
• Similarweb for competitors: monthly visits and channel mix for the domains you're up against. ahrefs.com logged about 8.3 million worldwide visits in August 2026 when we checked.

These ChatGPT prompts for keyword research cover two common setups. For a blog planning next quarter's content:

@firecrawl For the seed term "home espresso machine" in the US, get rising
related queries from Trends for the past 3 months and the past 12 months.
Group them into three clusters by intent and suggest one article title
per cluster.

For a local business:

@firecrawl Compare interest in "emergency plumber" and "24 hour plumber"
across US metros in Texas over the past 12 months. Tell me which phrase
leads in Houston, Dallas and Austin.

Each prompt sets the seed, the place and the window, which keeps the Trends calls narrow and the answer tied to data.

Pick between two phrasings

@firecrawl Compare "standing desk" and "walking pad" interest in the US
over the past 5 years and tell me which months peak for each.

Read the competition

@firecrawl Search for "best standing desk", scrape the top five pages and
list each one's H2s, word count and whether it has a comparison table.

Size up a competitor

@firecrawl Get worldwide monthly visits for flexispot.com and upliftdesk.com
for August 2026 from Similarweb, and show the price first.

Catch a trend early

@firecrawl What's trending in US searches over the past 48 hours? Show
the terms with approximate volume and related searches.

List tools and prices

@firecrawl Which tools can show search interest, related queries or site
traffic? List each with its price. Don't run anything.

Checking which tools exist and what they cost is free. Credits leave your Firecrawl balance only when ChatGPT runs a tool, and Firecrawl pricing covers plan allowances.

Each run ends with ChatGPT naming the tool and the credit charge.

• How to use ChatGPT for research: back up a content plan with papers and official statistics.
• How to use ChatGPT at work: competitor, sales and market research with other Firecrawl Alexandria providers.
• How to find podcast sponsors in ChatGPT: see which brands in your niche pay for podcast ads.

• Relative scores, not volumes. Trends reports interest on a 0 to 100 scale relative to the peak in your chosen range, so it ranks keywords against each other but doesn't give monthly searches. Low-volume terms can return empty lists.
• No difficulty score. None of these tools estimates how hard a keyword is to rank for. Judge that from the pages you scrape and the size of the sites behind them.