In short
- On a catalog of thousands of pages, the choice of production method matters more than the choice of words.
- Variable templates are fast but produce repetitive tags, which Google often rewrites.
- Manual writing gives the best results page by page, but its cost limits the scope to the main pages.
- Controlled AI combines volume and page-by-page treatment, provided it starts from real data and is proofread.
Optimizing the title tag of one category is simple. Optimizing those of 2,000 categories, facets and brand pages, in three languages, is an organizational problem. There are three main methods, each with its strengths.
Method 1: variable templates
The principle: a formula applied to all pages of the same type. "{Category name}: {number} models | My Store".
Strengths: instant, no writing cost, easy to maintain.
Limits: all pages look alike. Google explicitly cites repetitive text across pages (micro-boilerplate) among the causes of title rewrites, and the Zyppy study found a higher rewrite rate on titles built from a single template. Above all, a template knows neither the real query of each page nor what makes it different.
Best reserved for pages with no ranking stakes.
Method 2: manual writing
An SEO copywriter analyzes each page, identifies its query and writes tailor-made tags.
Strengths: the best possible quality page by page, when the copywriter is experienced.
Limits: cost and time. A few minutes per page for a title tag and a meta description, more with copy: on 2,000 pages, that is several weeks to several months. In practice, you optimize the 50 or 100 main pages, and the rest keeps the CMS tags. Quality also varies with fatigue and with copywriters.
Method 3: AI with quality control
A tool generates tags for each page, from its real data, then checks them.
Strengths: the page-by-page treatment of method 2, at the volume of method 1.
Limits: quality depends entirely on what the tool knows about each page and the checks it applies. A generator that only receives the category name will produce the equivalent of a template, at a higher cost.
Comparison
| Templates | Manual writing | Controlled AI | |
|---|---|---|---|
| Speed | Immediate | Slow | Fast |
| Cost per page | None | High | Low |
| Page-level specificity | Low | High | High if the tool reads the page |
| Risk of Google rewriting | High | Low | Low if the rules are good |
| Realistic scope | The whole catalog | Main pages | The whole catalog |
The criteria for a good generation tool
If you choose the third path, check that the tool:
- reads the real content of each page (products, brands, attributes, breadcrumb);
- relies on real queries (Search Console) and assigns one query per page;
- applies rules by page type: category, facet, brand, product page;
- proofreads what it produces and checks claims;
- publishes cleanly, with preview and rollback.
This is the grid against which unblind was designed. Its writing and proofreading rules are described in our title tag guide and our meta description guide.
Frequently asked questions
Can templates be used for SEO tags?
Yes, for secondary pages with no stakes (paginated pages, very similar product pages). For categories and facets that need to rank, a template produces interchangeable tags that Google often rewrites and that give no reason to click.
How long does it take to write a catalog's tags by hand?
Expect a few minutes per page for a quality title tag and meta description, including query analysis, and much longer with copy. On 2,000 pages, that represents several weeks to several months of work.