Bulk Title Tag and Meta Description Generation
A template that stamps the same phrase across five hundred pages produces five hundred titles Google is likely to rewrite. Ivan Sp generates title tags and meta descriptions across large page sets, each one matched to its own page content. One engagement produced recommendations for up to 500 pages on a single site, and another produced them per individual SKU across three sub-brands.
The Problem With Both Manual and Template Approaches
Writing 500 title tags by hand is possible and rarely finished. Sites in this position usually have hundreds of pages carrying a theme default or nothing at all.
A pure template produces the opposite failure. Concatenating a page name with a site name generates 500 near-identical tags that Google frequently overrides with its own text, which defeats the exercise.
What Generation Reads From Each Page
Generation works from the page’s own content rather than from its URL. The generator reads the visible headings, the body copy, and any structured data already present, then writes a tag describing what that specific page covers.
Consistency gets applied at the level of pattern rather than wording. A catalogue keeps a recognizable shape across its tags while each one names its own product, attributes, and category.
Constraints Applied to Every Output
Each generated element gets checked against a fixed specification:
- Title tags use the available character budget rather than stopping short of it
- Every title segment reads as a natural phrase rather than a string of keywords
- Meta descriptions run exactly two sentences, with the most important claim first
- Descriptions stay at a grade 6 reading level and remain factual rather than sensational
- No promotional superlatives appear unless the page substantiates them
Unused character budget counts as a defect rather than a safe choice. A title leaving 30 characters unused has given away relevance for nothing.
Entity Accuracy at Scale
Bulk generation multiplies naming errors. One draft title shortened a product name from “Capital Complete Health and Wellness Plan” by dropping two words, which changed the entity rather than tightening the copy.
Product and service names therefore get treated as fixed strings the generator cannot abbreviate. Where a name is genuinely too long for a title, the surrounding words get cut instead.
Templated Variables for Multi-Location Sets
Location page sets use variable placeholders rather than separate generation runs. One batch produced 200 image alt text attributes built around a city variable, each written to a strict specification covering length, sentence count, and excluded commercial words.
The same approach applies to titles and descriptions across a location set, which keeps a city set consistent while each page names its own market.
Strategy Compared to Production
This page covers producing metadata at volume. Deciding what a single high-value title should say, including whether a modifier helps or hurts the broader term, is a different question covered under title tag strategy for individual pages.
Where the Standard Bulk Metadata Advice Breaks Down
How many of your titles is Google currently displaying as you wrote them?
- The usual advice
- Build one title template with variables, apply it site-wide, and every page gets a unique title.
- What holds up in practice
- Variable substitution produces technically unique titles that are structurally identical. Google rewrites titles it judges boilerplate, so a site-wide template routinely produces five hundred titles that never display as written.
Consider a template producing the same phrase with one variable across five hundred pages. Every title is technically unique and the deployment takes an afternoon.
Google displays its own headline on most of them. The pattern reads as boilerplate, which is one of the documented rewrite triggers.
Ivan Sp generates from each page’s own fields rather than from one phrase. Titles then differ structurally as well as textually.
A rewritten title means the copy you deployed is not the copy anyone reads. The work shipped and the searcher never saw it.
Every rewritten title is a headline chosen by a machine instead of by you, on a page you already earned the ranking for.
Four Things Commonly Assumed About Bulk Metadata
Which of these describes your current metadata?
I meet these four assumptions on most metadata generation work.
| Commonly assumed | What actually holds |
|---|---|
| Variable substitution creates unique titles. | Substitution produces textual uniqueness and structural sameness. Google reads the pattern as boilerplate and rewrites it. |
| Shorter titles are safer at scale. | Unused character budget is relevance given away. Front-loading the differentiator matters more than cutting length. |
| Meta descriptions can be skipped at volume. | An empty description guarantees a composed snippet. A generated two-sentence description at least gives Google something matched to the page. |
| One template can serve every page type. | Different page types earn different queries. A product template applied to categories produces titles matched to neither. |
| Uniqueness only needs checking within a template. | Duplicates appear across templates too. The check has to run across the whole generated set. |
Titles are the copy a searcher reads before deciding to click. Generating them badly wastes the ranking you already paid to earn.
A Worked Example: Two Generation Approaches
Consider 500 pages given titles two ways. The comparison below is illustrative, written to show what field-driven generation changes. No client engagement is described here.
| Property | Single template with a variable | Field-driven generation |
|---|---|---|
| Title structure | Identical across all 500 | Varies with the fields present |
| Differentiator position | After the brand phrase | Within the first 55 characters |
| Character budget used | About half | Most of it |
| Descriptions generated | None | Two sentences per page from real fields |
| Duplicate titles across the set | None textually | None |
| Titles Google displayed as written | A minority | A large majority |
I generate from fields rather than from a phrase. The final row is the only measure that matters, and the first four rows are why it moves.
A single title template is a name badge printed with the same greeting for every guest. Technically personalised and instantly recognisable as a template.
The character budget row compounds. Half a budget unused across five hundred pages is a great deal of relevance left unclaimed.
How Bulk Generation Runs
I generate from each page’s own fields, then verify against a crawl rather than assuming.
- Pull the existing query set per URL from Search Console where the pages already have history.
- Map which data fields are available per page type, since availability varies across a set.
- Write a generation rule per page type rather than one rule for the whole site.
- Front-load the differentiating fields into the first 55 characters of every title.
- Use the remaining character budget on a second differentiator rather than leaving it empty.
- Place the brand name last, so truncation costs the least valuable words.
- Generate two-sentence descriptions from fields specific to each page.
- Run a uniqueness check across the entire generated set, not per template.
- Deploy, then crawl and export every title and description with its character count.
- Search a sample in Google after 28 days, and record how many titles display as written.
How This Engagement Is Scoped
I scope bulk metadata by page type count rather than by page count.
- What arrives
- A generation rule per page type, the generated set, and a post-deployment crawl report.
- Typical duration
- Two to four weeks for a set of several thousand pages.
- What the client supplies
- Field access, template access, and Search Console access.
- Included
- Query research, rule writing, generation, uniqueness checking, and verification.
- Not included
- Any promise that Google will display the generated titles.
- What changes the estimate
- Many page types, or a data source with inconsistent field availability.
- How it is quoted
- After a short scoping call, once page types and field availability are known.
Tools and Primary Sources
The references below define the display behaviour and the limits. Each opens in a new tab.
- Google title link documentation explains how the results headline is chosen and what triggers a rewrite.
- Google snippet documentation explains how the description snippet is selected.
- Search Console Performance report supplies the query history that field-driven generation should be matched against.
- Screaming Frog SEO Spider exports every generated title and description with length and duplicate flags.
Check How Many of Yours Get Rewritten
I suggest a check that takes about ten minutes.
- Search Google for site: followed by your domain.
- Read the first twenty headlines shown in the results.
- Compare each against the title tag in your own CMS or crawl export.
- Count how many differ.
Several rewritten headlines usually means the titles read as boilerplate to Google. The cause is structural sameness rather than length.
Common Questions
Are bulk-generated tags implemented or delivered as recommendations?
Usually delivered as recommendations, because applying 500 changes is its own scope. Where a system already generates the pages, the metadata publishes with them.
How do you check quality across hundreds of outputs?
A sample gets reviewed by hand, and the full set gets checked programmatically for length, duplicate tags, missing entities, and banned phrasing. Duplicates across the set are the most common fault.
Will Google use the generated titles?
Google rewrites titles when it judges its own version more relevant, which happens more often with templated and keyword-stacked tags. Natural phrasing specific to the page reduces the rate.
Why does Google rewrite generated titles more often?
Generated titles frequently share one structure across many pages, which reads as boilerplate. Generating from each page’s own fields produces structural variation as well as textual variation.
Discuss This Work
The first call is a free 30-minute consultation with no pitch attached. You receive an honest assessment of where the site stands and what can realistically be built.