bdos-ai-extensions

Group a flat list of keyword ideas into ad-group-ready clusters. Use after bdos-keyword-research when the user has 100s of Keyword Planner ideas and wants them organized into themed ad groups with rolled-up volume/CPC/competition and a suggested Ads structure. Three tiers — lexical (stdlib, zero install), fuzzy (rapidfuzz), semantic (embeddings + whitened-cosine threshold via an isolated heavy venv). Read-only.

DataForSEO API client for BDOS. Use when the user wants keyword search volume + CPC + competition (Google Ads / Keyword Planner data), keyword ideas/suggestions/difficulty, search intent, keywords for a site, ad-traffic estimation at a bid, SERP results, SERP competitors, Google autocomplete, Google Trends seasonality, Google Ads Transparency (competitor ad creatives), or Google Shopping / Merchant products & sellers. A thin REST client — independent of the dfs-mcp MCP server — that reads/analyzes only (never mutates a Google Ads account).

Advanced n-gram waste analysis of Google Ads search terms → negative keywords. Use when the user wants an n-gram / ngram breakdown of search terms with wasted-spend scoring (nScore), Cost Savings, Conv. Loss, Blocked Keywords/Search Terms, vs-average deltas, optional GA4 engagement columns, and concrete negative-keyword recommendations. Richer than the core search-terms n-gram: it ranks fragments by wasted spend and proposes negatives.

Profit-driven bidding decisions using marginal ERS and price elasticity (the "Zero-ROI model"). Use when the user asks whether to scale a campaign up or down, what target ROAS/tROAS maximizes PROFIT (not ROAS), whether extra spend is still profitable, price elasticity of traffic, marginal ERS/ROI, or "is raising bids/budget worth it". Pure Python math, no deps, no MCP.

Audit a landing page for Google Ads quality/relevance signals AND review it as sales copy for conversion. Use when the user wants to check a landing/destination URL for a campaign — technical signals (title, meta description, H1, mobile-friendliness, structured data, image alt, CTAs, thin content) plus a sales-copy/conversion review against copywriting frameworks (AIDA, PAS, FAB, value proposition, social proof, urgency, single clear CTA, objection handling). Technical audit is pure standard library and runs fully offline.

Offline competitor content comparison & content-gap analysis (pure Python, no MCP, no APIs). Use when the user wants to compare their page against competitor pages, check keyword coverage on a URL, find which competitor covers a topic best, or spot content gaps (keywords/topics missing from a page). Fetches pages with the standard library only — works fully offline.

Watch web pages for changes on demand — competitor, price and promo monitoring. Fetches a page, extracts readable text (markup stripped), stores a timestamped snapshot, and produces a unified diff against the previous snapshot. Pure standard library, no MCP, no pip deps, runs fully offline. Use when the user wants to track changes on a competitor page, detect price/promo updates, snapshot a page, or diff a page against the last time it was checked.

Extract and validate schema.org structured data (JSON-LD) from a web page, with a focus on Google Merchant Center / free-listing product eligibility. Use when the user wants to check a product page's structured data, find missing Product markup fields (name, image, price, brand, sku/gtin, availability), audit JSON-LD, or verify a landing page is "merchant ready" before running Shopping / free listings. Pure Python, no MCP, works offline.

Local URL / link health checker for Google Ads final URLs, sitelinks and landing pages (no MCP, pure standard library). Use when the user wants to verify that ad final URLs resolve to a healthy 200, inspect a redirect chain, catch https->http downgrades, batch-check many URLs, or crawl a landing-page domain for broken internal links. Runs fully offline — no pip deps, no venv.

Local web crawling, scraping and structured extraction via Crawl4AI (no MCP required). Use when the user wants to scrape a page, crawl a site, fetch a page as markdown, deep-crawl multiple pages, or extract structured data (prices, products, listings) from a website. Runs fully offline in a dedicated venv.