Competitor Teardown

Outcome: a repeatable brief on any competitor covering their paid strategy, traffic, keywords, content and positioning — five sources, one run, about 29 credits for five competitors.

Surface: MCP server, or app columns
Level: Beginner
Uses: b2b_ads_search · meta_ads_analysis · semrush_insights · similarweb_insights · linkedin_page_posts
Credits: ~29 for 5 competitors
Prerequisite: Lessons 05–07

Why this is worth a lesson

Competitive research is normally an afternoon of tab-switching, done once, and stale within a quarter. As a chain of five actions it takes minutes and re-runs whenever you want.

It is also the best possible demonstration of chaining: no single source tells you much, and the five together tell you what a competitor is actually doing.

The five sources

Source	What it tells you	Tool
B2B ads	Which messages they are paying to put in front of buyers — their tested positioning	b2b_ads_search
Meta ads	Creative, audience and how long each ad has run. Long-running ads are winning ads.	meta_ads_analysis
Keywords	What they rank for and pay for — where they get demand	semrush_insights
Traffic	Volume, trend and channel mix — how big and which way they are going	similarweb_insights
LinkedIn posts	What they are announcing, hiring for and celebrating	linkedin_page_posts

The chain

Start from the domain

Everything keys off it. One competitor at a time.

Ads first — paid says what they believe

Run the B2B ad search and the Meta ad analysis. Ads are the highest-signal source because they cost money: nobody runs a message for six months that does not work.

Then keywords and traffic

Keywords say which problems they are chasing. Traffic says whether it is working, and the trend matters more than the number.

Recent LinkedIn posts give you announcements, launches and hiring themes — the current quarter’s priorities.

Then synthesize

An AI step over all five outputs, producing a constrained brief rather than a dump.

The synthesis prompt

Raw output from five sources is not a brief. Constrain it.


You are given ad copy, keyword data, traffic data and recent social posts
for a competitor.

Produce exactly these sections, no preamble:

POSITIONING: one sentence — how they describe what they do
PRIMARY BUYER: one of SMB, Mid-market, Enterprise, Mixed, Unknown
TOP 3 CLAIMS: three bullets, taken from their ads verbatim where possible
TRAFFIC TREND: one of Growing, Flat, Declining, Unknown
CURRENT PUSH: one sentence on what they are emphasising this quarter
WHERE WE DIFFER: two bullets, based only on the evidence above

If a section cannot be answered from the data, write Unknown.

“Taken verbatim where possible” matters. A competitor’s own ad copy is far more useful than a paraphrase — it is the wording their buyers have already been trained on, and it tells you exactly which objection your own message needs to answer.

Running it as a prompt chain

From an MCP client, this is one conversation. A prompt that works:


Using the SyncGTM MCP server, build a competitor brief for acme.com:

1. Search B2B ads for their domain and summarize the messages.
2. Analyse their Meta ads — note any ad running longer than 60 days.
3. Pull SEMrush insights: top organic and paid keywords.
4. Pull Similarweb insights: monthly visits, trend, channel mix.
5. Pull their last 10 LinkedIn page posts.

Then produce the brief in the format below.
[format block]

Same chain as columns in a table if you want it for many competitors at once — see AI for Sales Reps for the chat surface and SyncGTM 101 for the table one.

What to do with it

Sales enablement. A battlecard built from what the competitor is actually saying this quarter, not last year’s deck.
Positioning. Where their claims cluster is where the category is crowded. Where nobody is claiming is where you can.
Demand. Keywords they pay for are keywords that convert for someone in your category.
Timing. A competitor pushing hard into a segment is a reason to defend your accounts in it.

Do this now

Pick one competitor

The one you lose to most, not the biggest name in the category.

Run the five sources

One at a time. Read each output before chaining.

Note the long-running ads

Anything live for more than 60 days is a tested, working message.

Run the synthesis

With the constrained format.

Check it against what you know

You already know this competitor. Where the brief disagrees with you, find out which is right — sometimes it is the brief.

Repeat for four more

About 29 credits for five, and you have the competitive set covered.

Schedule a re-run

Quarterly. Positioning changes, and the change is the interesting part.

Check your work

All five sources returned data, or you know why one did not
The brief follows the format, with Unknown where data was missing
Claims are quoted from real ad copy rather than paraphrased
You have briefs for your whole competitive set, not just one
A re-run is scheduled

Where this breaks

Ad and traffic tools cover large advertisers well and small ones patchily. A competitor running no paid ads returns nothing — which means “no ads found”, not “no ads”. Treat empty results as missing data rather than as a finding, and mark those sections Unknown. A brief that concludes a competitor has no paid strategy when the tool simply has no coverage will lead a sales team somewhere unhelpful.

Further automation

Re-run quarterly and diff against the previous brief. The change is the signal: new claims mean new positioning, a traffic decline means pressure, a hiring push in a region means expansion. A standing competitor watch is one of the cheapest recurring research jobs you can run, and Signals & ABM covers turning the diff into an alert.

Next lesson

09 — Research at scale and what it costs, taking these techniques from 50 rows to 5,000 without the bill following linearly.

Reference for this lesson: b2b_ads_search, semrush_insights, similarweb_insights, linkedin_page_posts.