Batch-Writing TDK with AI: A Reusable Prompt Structure

The more structured your input, the less templated the output. Six required inputs, five hard constraints and one variation rule to avoid site-wide sameness.

Batch-Writing TDK with AI: A Reusable Prompt Structure

Many people use AI for TDK and few produce anything usable. The usual symptoms: one sentence pattern site-wide, descriptions built from adjectives, keyword stuffing. The model is not the problem; the input is too casual. Structure the input and output quality shifts noticeably.

Six required inputs

  1. Site name and URL. 2. Site structure, meaning which sections exist. 3. Main product categories. 4. Target markets by country or region. 5. Target customer types: wholesalers, importers, OEM buyers, contractors. 6. Page specifics: full product name, material, specification, application, certifications. Miss one and output degrades into a generic template.

Five hard constraints

First, titles of eight to twelve words separated by pipes or underscores, with the product name in the first part and marketing modifiers plus keyword variants in the next two. Second, the description's first sentence must contain the product name, later sentences weave in variants, total length 140-160 characters. Third, no keyword stuffing and no repeated term within a sentence. Fourth, output in both English and Chinese with lowercase English keywords. Fifth, modifiers must not repeat across different products.

One variation rule for switching products

This is the most overlooked and the most effective against templating: each time you change product, require new phrasing and new modifiers, with different sentence structures for title, description, H1 and H2. Replacing only the product name is not acceptable. State it explicitly: the previous product used modifiers A, B and C, which must not appear again.

A copy-ready prompt

Product: full English name. URL: link. Product information: material, appearance, specification, packaging, certification. Target market: country. Target customers: type. Tasks: 1) write a title of 8-12 words containing the core keyword; 2) write a meta description whose first sentence contains the product name; 3) produce lowercase English keywords; 4) provide both English and Chinese; 5) keep the title at 50-60 characters and the description at 140-160; 6) do not reuse modifiers from the previous page.

Three things humans must still do

First, verify facts: materials, specifications and certifications cannot be invented. Second, check real differentiation: read ten product TDKs in sequence and rewrite if the patterns match. Third, search-validate: run the core term through search and confirm the page delivers what the title promises.

Balancing speed and quality

The value of AI is compressing one hundred pages of TDK from two weeks to two days, not turning well-written into fast-written. The sensible workflow is AI drafts, human fact-checking and differentiation, then one style pass. All three steps together give fivefold throughput without a quality drop.

Compiled by the aibolg team from project practice and public industry material, for reference only.
← Back to listContact us →

Related Articles