Video summary

Генерируем тексты для SEO с Claude + ChatGPT в 2026 году за 30 минут

Main summary

Key takeaways

Educational

Main ideas / lessons

  • Goal of the process: Generate high-quality SEO text for existing commercial landing pages so they:
    • rank in Google/Yandex,
    • attract search engine traffic for the correct keywords,
    • support generative AI / GMI / Google AI Overview recommendations.
  • Core method: Create a technical assignment (spec) with structured requirements (headings, keywords, volumes, placement rules). Then use neural networks (ChatGPT + Claude) to turn the spec into a draft, followed by human post-editing.
  • SEO heading strategy: Use a controlled heading hierarchy—especially H2 and H3—to emphasize key phrases that might otherwise be missing.
  • Keyword distribution principle: Don’t overload headings with too many keywords; embed many keywords naturally inside paragraphs.
  • Page UX/readability matters: Use formatting such as bulleted lists, sensible paragraph sizes, and avoid visually “thin” layouts.
  • Interlinking is part of the solution: Cross-link related pages to form a web-like internal structure, helping indexing and distributing page weight.
  • Include commerce signals + credibility factors: Add commercial intent trigger words, precise numerical data, and author/expert opinion to make the text more useful for users and robots.
  • Use prompts responsibly: Free versions of GPT/Claude can work, but prompts must explicitly request text (not code) and specify required formatting and post-processing.
  • Final step is mandatory: After AI generation, run checks (originality/spam/roboticity) and do manual cleanup (logic, formatting, heading-level corrections).

Detailed instruction / methodology (step-by-step)

A) Build the Technical Assignment (inputs before generation)

Include fields such as:

  • Page identifiers:
    • page type (commercial landing),
    • page title,
    • “seller” (service/provider).
  • Heading structure requirements:
    • H1 already exists on many sites; focus especially on H2 and below.
    • Use H2 for major emphasis on top keywords.
    • Use H3 under each H2 (maintain hierarchy).
  • Keywords table:
    • For each phrase:
      • frequency (global scope if targeting all Russia),
      • competitiveness level (1 = low, 2 = medium, 3 = high),
      • priority (used to decide where the phrase goes in the text).
    • If a site has little/no traffic, prioritize low-competition keywords first (competitiveness = 1).
  • Placement rules:
    • Some phrases go into H2/H3 based on priority.
    • Lower-priority phrases go into paragraphs (e.g., within “div”/body text).
  • Text volume target: example given ~2000 characters.

Formatting/URL rules for headings/paths (technical constraints):

  • Avoid underscores.
  • Avoid relying on automatic uppercase/lowercase conversions; ensure correct casing.
  • Avoid numbers in user-friendly URLs (preference/tradition).

B) Use the “7 SEO/Geo/Generative” checklist for strong landing-page text

  1. Display all semantics: cover the keyword set from the spec.
  2. Emphasize commercial intent with trigger terms, e.g.:
    • order, availability, purchase/buy, in stock, online store, rent, sale, etc.
  3. Include precise numerical data:
    • internal stats, research metrics, concrete numbers.
  4. Add author/expert opinion:
    • a quote from an author/team member,
    • an opinion paragraph (not only facts).
  5. Implement interlinking (internal links):
    • build a “spider web” of links across the site,
    • usually link within sections (parent category → related pages),
    • anchor text should support cross-linking.
  6. Geographic binding (only if needed):
    • if serving a region, add a city-specific paragraph (e.g., Moscow),
    • if serving Russia broadly, avoid over-binding to a city; follow Russia/CIS strategy.
  7. Bonus for generative/geo traffic (Q&A block):
    • create a Q&A module under the text,
    • include questions users/AI commonly ask (how much, where, what is better, top, advice, etc.),
    • ensure answers are specific and include digital data.

C) Generate the draft using two neural-network prompts

Use two prompts:

  • Prompt 1 for ChatGPT (GPT):
    • Assign a role to the GPT “developer”.
    • Ask it to convert dry tabular specs into a structured, detailed technical text/spec suitable for Claude.
  • Prompt 2 for Claude (Claude):
    • Provide the structured spec/template.
    • Explicitly instruct: generate text, not code (to avoid misunderstandings).
    • Request required formatting and mention post-processing needs (removing/adjusting AI-added markup).

The speaker notes free versions can help reduce cost, and also mentions intermediary services as an option.

D) Post-generation corrections (human-in-the-loop)

Validate and fix:

  • Heading level mistakes: Claude may shift some H2 terms into H3 unexpectedly.
  • Remove unnecessary introduction if it doesn’t fit a landing page.
  • Adjust titles (e.g., “Conversion growth 24x7” should be turned into a proper title phrase).
  • Ensure required quoted/opinion content exists.

Add missing elements:

  • If expert opinion is missing, add a quote paragraph (example: “Ivan” as an expert).

Improve SEO/copy quality:

  • Bold key phrases that correspond to technical spec phrases.

Run copy checks:

  • Use a tool (e.g., text.ru) to check:
    • originality,
    • spam/over-spam,
    • robotic/nonnatural patterns.

Insert into CMS/admin correctly:

  • ensure correct H1/H2/H3 placement and overall structure.

E) Final workflow output

  • The AI produces a strong starting draft.
  • A copywriter/SEO specialist finalizes:
    • manual tweaks,
    • logical cleanup,
    • completeness and correctness against the spec.

Speaker / sources featured

  • Speaker: Unnamed host/speaker from the WebUS channel (mentions personal team, SEO work, and personal expertise).
  • Tools / platforms mentioned:
    • ChatGPT (GPT)
    • Claude
    • Wordstat
    • Yandex
    • Google
    • Google AI Overview / GMI / “generative search results”
    • text.ru (for originality/spam/roboticity checking)
  • Named person referenced in the example: Ivan (used as the expert for author opinion).

Original video