AI-powered SEO keyword research with live search trends

SEO keyword research with live search trends is a ranking of candidates by slope: pull Google Search growth for a list, keep terms with strong 1 month and 3 month change while the index is still modest, then check Shopping, YouTube, and history before anyone outlines a page. As of August 2026, one MCP or REST API covers 25+ platforms, stores weekly history of about five years, and scores series on a 0-100 index. The free plan is 100 requests per month. Starter is $19 per month. Difficulty and current volume still come from a keyword suite. Timing comes from the trend series.

Ranking keywords by slope instead of size

Keywords are ranked by slope because a mid-volume term that is compounding beats a giant term that has been flat for three years, and by the time a query is fat in every planner it is also fat in every rival's planner, which makes the window the still-cheap term that is already bending up. The window is the term that is still cheap and already bending up.

A practical pass is a spreadsheet of 20 to 50 candidates from sales calls, Reddit phrasing, and old rank trackers. Each row gets get_growth on Google Search for 3 months and 1 year. Strong 1 month plus 3 month with a still-low index is the early target. A huge 1 year number with a negative 3 month number is a peak that already happened. Ranked Google Search lists and live Google Trends feeds add names that were not on the sheet yet.

Weekly history of about five years is the seasonality filter. "Best CRM software" in January is not the same decision as the same phrase in August. The 0-100 index shows whether this winter is higher than last winter, which a single monthly volume cell will not.

How a candidate list gets scored in one session

A candidate list gets scored in one session when the assistant receives the list, calls growth for each term, and returns a table sorted by 3 month change with the 1 year change beside it, after which humans drop terms that the site cannot serve regardless of slope. A medical domain should not chase a rising slang query it cannot legally cover.

get_trends on weekly mode is the second gate for anything that looks attractive. A series that rolled over last year can still have leftover volume. Publishing into a decline wastes the outline. Re-scoring the same list each month is how a program stays honest. Content strategy consumes that table as calendar order. SEO consumes it as targeting order.

Clusters form when related terms move together on Search and Shopping. The assistant can query a parent term and a few modifiers in the same sitting. If "CRM" is flat and "AI CRM" is climbing, the cluster has a new head. That is a page brief, not a synonym tweak.

When Shopping or YouTube disagrees with Search

Shopping and video series are intent checks on the same keyword, not extra vanity charts, because a term growing on Google Shopping while Google Search is quiet often means purchase intent is ahead of research intent, which favors comparison and product pages. A term growing on YouTube while Search is quiet often means people want a demonstration first.

Amazon search trends add a retail check for affiliate and ecommerce briefs. Amazon climbing before Google Search is a common sequence for product categories. The SEO page can still win the later informational queries if it ships during that lag. Teams that only watch Search miss the buy-side move.

Google Trends API coverage on this stack is the programmatic Search series with a shared index across other sources. Official Google Trends in the browser remains a relative chart with no public API. The research workflow that needs Shopping, YouTube, and Search in one JSON packet is the MCP or REST path.

Using difficulty tools with live momentum

Ahrefs and Semrush still own difficulty scores, ranking URL lists, and estimated current volume, while a multi-source trend API that covers 25+ platforms, weekly history of about five years, and a 0-100 index owns trajectory and cross-platform confirmation, so the two sit in sequence with momentum first and difficulty second. Complementary means both get used. A rising keyword with KD that the domain cannot touch is a long-term target or a different format, not this week's post. A low-difficulty keyword with a collapsing weekly series is a trap. Keyword suites do not replace the slope. The slope does not replace the SERP.

Vendor pricing for Ahrefs and Semrush is public on their sites and changes. Check current pricing rather than copying an old seat cost into a brief. As of August 2026, trend queries start at 100 requests per month on the free plan, with Starter at $19 per month, which is enough to score a modest candidate list weekly if terms are not re-pulled every hour.

Timing content to the growth window

Content is timed to the growth window by shipping while 3 month growth is still positive and the weekly series has not already peaked, then refreshing when the slope flattens instead of when traffic finally drops, because waiting for enough volume in a planner is how teams arrive after the SERP is full.

Honest limits: a 0-100 index is not a keyword planner volume. Absolute estimates, where present, are directional. News-driven queries can spike and die in two weeks. Five years of weekly points help, but a brand-new slang term will have a short series. Rank tracking, crawl data, and conversion data remain separate systems.

The assistant can draft titles from the ranked table. It cannot promise position one. The tools below are the calls that scoring loop uses.