Skip to content

Discount-Code & Affiliate Sales Mastery · AI-assisted outreach and automation · lesson 12 of 12 · 16 min

AI-assisted outreach, pitching and content at scale

Where AI genuinely helps an affiliate seller

AI tools such as ChatGPT, Claude or Gemini are useful across the affiliate workflow, as long as a human stays accountable for every message and every claim:

| Task | Good use of AI | Keep human | |---|---|---| | Brand prospecting | Summarise a brand's public site, products and past creator campaigns | Deciding fit (use your scorecard) | | Pitch emails to brands | Draft a personalised first version from your notes and stats | Facts, numbers, tone, sending | | Replies to followers' questions | Draft answers from the brand's FAQ and your notes | Anything about health, money or safety; final wording | | Content | Structure scripts, captions, comparison tables from your testing notes | The experience itself; approved claims | | Reporting | Summarise your sales log, spot patterns | Numbers you submit | | Translation | First drafts in Urdu, Arabic or other languages | Native-speaker review |

The rules that still apply when AI writes

  • Honesty and consumer law. AI-written reviews of products you have not used, invented statistics, and fake testimonials are misleading. In the US, the FTC's rule on consumer reviews and testimonials bans fake reviews including AI-generated ones; the UK bans fake reviews under the Digital Markets, Competition and Consumers Act 2024.
  • Disclosure. AI-drafted captions still need "Ad" and commission disclosure upfront.
  • Marketing messages. In the UK, PECR generally requires consent before sending marketing emails or texts to individuals (with a limited "soft opt-in" for your own customers); business-to-business emails to corporate addresses are allowed but must identify you and offer an opt-out, and UK GDPR still applies. In the US, CAN-SPAM requires accurate headers, a clear opt-out and a physical address in commercial emails. The EU, UAE and KSA have their own data protection and messaging rules. When in doubt, message only people who asked to hear from you, or businesses whose public contact route invites partnership enquiries.
  • Platform rules. Platforms restrict automation. LinkedIn's terms prohibit bots and automated messaging or scraping; Meta's messaging on Instagram and WhatsApp must go through official APIs, with opt-in and template rules for WhatsApp Business. Automating DMs with unofficial tools risks your account.
  • AI chat agents. If you run an AI assistant that answers customers, tell people it is AI. For audiences in the EU, Article 50 of the EU AI Act requires this from 2 August 2026.

Worked example: pitching brands with AI, honestly

Bilal, a Birmingham tech creator with a strong UK and Pakistani audience, wants three new programmes a quarter. He keeps a prospect sheet with public research notes, uses AI to draft each pitch from his own verified stats, edits every draft, and sends from his own inbox to each brand's published partnerships address. Illustratively, his reply rate improved once pitches referenced each brand's actual products and his real earnings-per-click history, and he spent less time per pitch.

Hands-on 1: a pitch prompt that cannot invent facts

You are helping me write a first-draft pitch email to a brand's
partnerships team. Use ONLY the facts below. If a fact is missing,
write [NEED: ...] instead of inventing it. British English. Under 150 words.
No flattery about posts I have not referenced. No promises of results.

ME: [name], [niche], [platforms + follower counts as of date]
MY RESULTS (verified): [e.g. avg EPC GBP 0.42 across 3 tech programmes, Q2 2026]
AUDIENCE: [countries, age bands, top questions they ask]
BRAND FACTS (from their public site): [products, markets, existing programme link]
WHY FIT: [2 reasons from my scorecard]
ASK: [join programme / test code / specific offer idea]

Hands-on 2: draft pitches in bulk, review every one

This script reads a CSV of prospects you researched, asks Claude to draft a pitch for each using only your facts, and writes drafts to a new CSV for human review. It never sends anything. Set ANTHROPIC_API_KEY in your environment; never paste keys into code.

# pip install anthropic
import csv
import os
import sys

import anthropic

MODEL = os.environ.get("ANTHROPIC_MODEL", "claude-opus-5")  # check current model names in Anthropic's docs
MY_FACTS = open("my_facts.txt", encoding="utf-8").read()     # your verified stats and audience notes

client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY from the environment

SYSTEM = ("You draft short partnership pitch emails. Use only facts provided. "
          "If something is missing write [NEED: ...]. Never invent numbers, quotes or past collaborations.")


def draft(prospect: dict) -> str:
    prompt = (f"MY FACTS:\n{MY_FACTS}\n\nBRAND RESEARCH NOTES:\n{prospect['research_notes']}\n\n"
              f"Write a pitch to {prospect['brand']} ({prospect['contact_route']}). Under 150 words, British English.")
    msg = client.messages.create(model=MODEL, max_tokens=4000, system=SYSTEM,
                                 messages=[{"role": "user", "content": prompt}])
    if msg.stop_reason == "refusal":
        return "[MODEL DECLINED - write manually]"
    return "".join(block.text for block in msg.content if block.type == "text").strip()


with open("prospects.csv", newline="", encoding="utf-8") as src, \
     open("pitch_drafts.csv", "w", newline="", encoding="utf-8") as out:
    reader = csv.DictReader(src)  # columns: brand,contact_route,research_notes,lawful_basis
    writer = csv.DictWriter(out, fieldnames=["brand", "contact_route", "draft", "status"])
    writer.writeheader()
    for p in reader:
        if not p.get("lawful_basis"):
            print(f"skip {p['brand']}: no lawful basis/contact permission recorded", file=sys.stderr)
            continue
        try:
            text = draft(p)
        except anthropic.RateLimitError:
            text = "[RATE LIMITED - retry later]"
        except anthropic.APIStatusError as err:
            text = f"[API ERROR {err.status_code}]"
        except anthropic.APIConnectionError:
            text = "[CONNECTION ERROR]"
        writer.writerow({"brand": p["brand"], "contact_route": p["contact_route"], "draft": text, "status": "needs human review"})

Review each draft against your facts, delete anything untrue, rewrite it in your voice, and send manually. Keep the sheet as a record of who you contacted and why.

Measuring success

  • Reply rate and programme acceptance rate per 10 pitches.
  • Share of AI drafts that needed factual corrections (should fall as your facts file improves; never zero checking).
  • Complaints or unsubscribes from outreach (target: zero).
  • Time per pitch.

Video lecture: AI-assisted outreach, pitching and content at scale

Lecture coming soon · 15 chapters · about 9 minutes. Read the full transcript below.

  1. AI-assisted outreach and content
  2. Why it matters
  3. The keen junior assistant
  4. Where AI helps
  5. Rules that still apply
  6. Messaging rules
  7. Platform automation rules
  8. Example 1: Lahore fashion creator
  9. Example 2 (illustrative): Bilal, Birmingham
  10. Watch me do it: fact-locked pitch
  11. Bulk drafting, safely
  12. AI translation
  13. Common mistakes
  14. Measure it
  15. Recap + try this now

Lecture transcript

AI-assisted outreach and content

You could send a hundred brand pitches this afternoon. An AI tool can write them all in a few minutes. So why do most AI-written pitches get ignored, and some get creators banned from platforms? Because speed isn't the hard part. Truth, relevance and permission are. In this lecture you'll learn where AI genuinely helps an affiliate seller, the rules that still apply when AI writes, a pitch prompt that can't invent facts, a script that drafts in bulk for human review, and how to measure whether it's working.

Why it matters

Why does this matter? Because AI multiplies whatever you feed it. Give it vague prompts and it produces generic flattery that brands delete on sight. Give it permission to guess and it invents numbers, which can end a relationship the moment a brand checks. Automate messaging carelessly and you break platform rules and marketing law. But use it well, and you can research brands faster, write sharper pitches, answer followers quicker and translate into more languages, while keeping your voice and your integrity.

The keen junior assistant

Here's the analogy. Treat AI like a very fast, very keen junior assistant who has never met your audience and sometimes makes things up to be helpful. You'd let them research, draft and tidy. You wouldn't let them sign contracts, quote numbers you haven't checked, or email strangers on your behalf. Same with AI. It drafts. You decide, verify and send.

Where AI helps

Where does AI genuinely help? Brand prospecting: summarising a brand's public site, products and past creator campaigns. Pitch emails: drafting a personalised first version from your notes and verified stats. Replies to followers: drafting answers from the brand's FAQ and your notes, but never on health, money or safety without your own careful review. Content: structuring scripts and comparison tables from your testing notes. Reporting: summarising your sales log. And translation: first drafts in Urdu or Arabic, checked by a native speaker.

Rules that still apply

Now, the rules that still apply. Honesty: AI-written reviews of products you haven't used, invented statistics and fake testimonials are misleading. In the US, the FTC's rule bans fake reviews, including AI-generated ones, and the UK bans fake reviews under its twenty twenty-four consumers act. Disclosure: an AI-drafted caption still needs Ad and your commission disclosure upfront. And if you run an AI assistant that answers customers, say it's AI. For audiences in the EU, that's a legal duty from August twenty twenty-six.

Messaging rules

Messaging rules matter even more when AI makes sending easy. In the UK, marketing emails and texts to individuals generally need consent, with a limited soft opt-in for your own customers. Emails to corporate addresses are allowed, but must identify you and offer an opt-out, and data protection law still applies. In the US, commercial emails need accurate headers, a clear opt-out and a physical address. The EU, UAE and Saudi Arabia have their own rules. The safe principle: message people who asked to hear from you, and businesses through the partnership contact they publish.

Platform automation rules

And platforms have their own rules. LinkedIn's terms prohibit bots and automated messaging or scraping. Messaging on Instagram and WhatsApp for business must go through Meta's official tools and APIs, and WhatsApp Business requires opt-in and approved templates for business-initiated messages. Unofficial DM automation tools might work for a week. Then the account you've spent years building gets restricted. It's never worth it.

Example 1: Lahore fashion creator

First example. A Lahore fashion creator gets dozens of DMs a day asking about sizing and delivery. She pastes the brand's FAQ and her own notes into an AI assistant and asks it to draft short replies, one per common question, in English and Urdu. She edits them, has a friend check the Urdu, and saves them as quick replies. She still sends each one herself. Her response time drops from a day to minutes, and every answer is accurate.

Example 2 (illustrative): Bilal, Birmingham

Second example, a realistic business scenario with illustrative numbers. Bilal, a Birmingham tech creator with UK and Pakistani audiences, wants three new programmes a quarter. He keeps a prospect sheet with public research notes and a column recording each brand's published partnerships address. He writes a facts file with his verified stats, uses AI to draft each pitch from only those facts, edits every draft, and sends from his own inbox. Illustratively, his reply rate roughly doubled compared with his old generic template, and each pitch took about ten minutes instead of forty.

Watch me do it: fact-locked pitch

Watch me do it. I open the pitch prompt from the lesson and fill in my facts: niche, platforms, follower counts as of this month, my average earnings per click across three tech programmes last quarter, my audience countries, and the top questions they ask. For the brand, I paste notes from its public website. I run it. The draft is good, but it says I previously worked with a competitor. I didn't. That's not in my facts, so I delete it. Another line says NEED: link to your best tutorial. Good, the prompt flagged the gap instead of inventing. I add the link, rewrite the opening in my voice, and send it to the brand's published partnerships address.

Bulk drafting, safely

If you have lots of prospects, the Python script in the lesson helps. It reads your prospects sheet, skips any row without a recorded lawful basis or contact permission, asks Claude to draft a pitch using only your facts, handles rate limits and errors, and writes every draft to a new sheet marked needs human review. Notice what it doesn't do: it never sends anything. Your API key lives in an environment variable, never in the code. And the model name comes from a setting, so you can update it when models change.

AI translation

Translation deserves its own mention, because it's one of the biggest wins for creators in Pakistan, the Gulf and the UK. AI can give you a solid first draft of a caption, FAQ answer or pitch in Urdu or Arabic in seconds. But machine translation can get tone, honorifics and product terms wrong, and a disclosure that reads oddly may not be understood. So treat AI translation as a draft. Keep a glossary of your key terms and approved disclosure lines, ask the model to use it, and have a native speaker check anything commercial before it goes out.

Common mistakes

Common mistakes. Letting AI invent numbers, past collaborations or experience. Sending drafts without reading them. Generic flattery that proves you didn't look at the brand. Mass messaging people who never asked. Using unofficial automation tools on social platforms. Pasting followers' personal data into consumer AI tools. And skipping native-speaker review of translations, which can turn a friendly message into an awkward one.

Measure it

How do you measure it? Track reply rate and programme acceptance per ten pitches. Track the share of AI drafts that needed factual corrections; it should fall as your facts file improves, but you never stop checking. Track complaints and unsubscribes from outreach, and aim for zero. And track time per pitch. Faster is only a win if replies and trust hold up.

Recap + try this now

Recap. AI is a fast, keen assistant: let it research, draft, structure and translate, while you verify every fact and send every message yourself. Lock prompts to your verified facts and make the model flag gaps. Respect consent, messaging rules and platform automation limits, and never let AI invent reviews or experience. Try this now: write your facts file, draft three pitches with the prompt from the lesson, correct them and send them manually to published partnership contacts. Then take the final assessment. Well done.

Key takeaways

  • Use AI to research, draft and translate, while a human stays accountable for every fact, claim and message sent.
  • Constrain prompts to your verified facts and mark gaps instead of letting the model invent them.
  • Outreach must respect consent and messaging rules (PECR, CAN-SPAM, data protection) and platform automation limits.
  • AI-written fake reviews or experiences are misleading and banned in key markets; AI chat agents must say they are AI.

Try it

Write your verified facts file, then use the pitch prompt to draft three brand pitches, correct every draft against your facts and send them manually to published partnership contacts.