Is programmatic SEO worth doing now?

What is programmatic SEO?

It uses structured information and a repeatable template to produce pages. Product specifications, service availability, comparisons or location-specific details can supply the variation. AI may help explain those records, but a language model is not the authoritative source of prices, coverage, stock or evidence.

A useful templated catalogue lets a customer compare real options or act on current information. Generic summaries are easier for an answer service to repeat without sending anyone to the site. A page therefore needs a reason for a relevant person to continue: verified compatibility, live availability or a useful transaction, for example. More published pages alone supplies none of that value.

Where does it create worthwhile value?

Look for repeatable questions with materially different answers. A buyer comparing compatible components may need actual dimensions and exclusions. A service buyer may need a locally different delivery cutoff, access requirement or available service. The distinction must matter to the decision, rather than merely change a town name in the introduction.

Check search and customer evidence for the intended page groups. Inspect representative results, existing queries and sales questions. Establish whether a separate page is a useful destination or whether one stronger page and a clear filter would serve the need better. Keyword tools can corroborate demand, but a blank estimate does not prove that no customers need the information.

Avoid competing pages for the same task. Decide the main page for each meaningful intent and connect it through an understandable site hierarchy. Relevant internal links, correct canonical choices and a controlled set of indexable URLs matter more than generating every possible filter combination.

What does Google's policy actually prohibit?

Google's scaled-content-abuse policy concerns generating many pages primarily to manipulate rankings without helping users, regardless of how the content is produced. Its doorway policy also addresses substantially similar regional pages that funnel users towards the useful destination. This is not a blanket prohibition on templates or AI assistance. Google's spam policies

Google's current AI-content guidance calls for accuracy and review, including page titles, descriptions and structured data. A convincing paragraph cannot compensate for an invented product fact. Google's AI-content guidance

A ranking fall alone does not establish a specific penalty. Assess the actual page family and applicable policy; no page count by itself establishes compliance or commercial value.

How should you screen the page plan?

Use an eligibility decision before text generation. Each candidate needs an actual customer task, a real available offer or information need, authoritative fields, a material difference from existing pages and a useful next step. Record the source owner, freshness requirement and behaviour when information is missing.

Separate candidates into pages worth developing, information better consolidated elsewhere, and combinations that should not be offered or generated. A missing value must not become an invitation for an assistant to invent a reassuring answer.

Consider this constructed service catalogue. Eight towns multiplied by six service types gives a proposed 48 pages. The business actually offers only twelve of those combinations. Thirty-six fail the offer check before writing begins. Of the remaining twelve, eight repeat information already adequately covered by existing service pages. Four have distinct, verified delivery or access conditions that customers need to resolve a separate question.

Those four are pilot candidates, not automatically four approved pages. Inspect whether a separate destination is still useful, check the facts and test the next step. The other eight may belong in existing coverage information. Publishing all 48 would manufacture unsupported availability and redundant content from a small underlying offer.

What does the cost per useful page reveal?

Assess the original, unrevised 48-draft proposal. It quotes generation at £120, or £2.50 each, and adds £480 technical setup in cash. Reviewing all 48 drafts takes 24 hours at 30 minutes each. At the assumed £50 hourly value, review consumes £1,200 of capacity before the team knows how much it can use.

If only the four distinct candidates ultimately meet the agreed acceptance standard, cash production cost is £600 divided by four: £150 per accepted page. Combined cash and capacity cost is £1,800 divided by four: £450. The initial £2.50 claim described draft generation, not usable publishing output. These values exclude ongoing maintenance and do not establish a return.

The sensible process screens the dataset first, so the team need not pay to write and review all the rejected combinations. Require suppliers to price data preparation, validation, integration, editing and maintenance alongside generation. More draft output can increase total cost when acceptance is poor.

Can the accepted pages earn enough to justify the commitment?

Keep those original proposal costs for this twelve-month illustration, with only four accepted pages. A properly screened four-page pilot needs a new quote; these are not its unavoidable setup costs. Assume monthly maintenance adds £80 cash and two hours at £50: £180 in combined resources. The annual cash commitment is £600 setup plus twelve times £80, or £1,560. The resource commitment is £1,800 setup plus twelve times £180, or £3,960. Use £110 contribution per genuinely additional retained order after identified variable fulfilment costs, before this programme's charges.

Fifteen additional orders would exceed the annual cash commitment. Thirty-six would exactly recover the combined resource commitment; thirty-seven would exceed it. Orders merely displaced from an existing service page do not count as additional business. The team can now ask whether the four useful pages can plausibly supply that demand, rather than treating eligibility as an investment case.

The forecast to replace with your own evidence is qualified visits across the eligible pages × retained-order rate × contribution per order, compared with the full period’s costs. Discovery, demand and timing determine the visits; none is guaranteed by page eligibility. For a deliberately hypothetical sensitivity, assume the four pages together attract 60 qualified visits monthly, with no displacement. At a 3% retained-order rate, expected monthly contribution is 60×3%×£110=£198. At 6%, it is £396. The annual resource commitment averages £330 a month. The lower case fails that requirement; the higher case exceeds it by £66 if it begins immediately and holds. These fractional expected order counts are forecasting averages, not observed purchases or conversion benchmarks.

Ramp-up changes the decision again. If the higher case earns for only nine of the twelve months, contribution is £3,564, below the £3,960 resource commitment. The same proposal can appear affordable on cash while failing on capacity or delayed demand. Inspect the assumptions using real search and customer evidence, compare an investment in existing pages and test which outcome is genuinely additional. None of these selected figures predicts the pilot's result.

How would you run the pilot?

Choose a small representative set from the eligible candidates, including awkward cases such as missing fields, changing availability and unusual restrictions. Freeze source records, page purpose, acceptance criteria and a maximum commitment before generating them.

Check the complete rendered pages and the generator together. Verify every consequential field, links, metadata, structured data and the intended main address. Inspect mobile use and the real enquiry or purchase route. Test that an updated source changes the relevant page, and that unavailable information triggers the agreed hold, consolidation or removal behaviour rather than fabricated copy.

After any authorised publication, check discovery and indexing, then relevant search exposure, qualified visits, outcomes and the maintenance burden over a suitable window. Indexing is neither guaranteed nor proof of useful demand. Compare the result with improving existing pages and other credible uses of the same resources.

What would justify scaling or stopping?

Expand only where distinct useful pages can be sustained with reliable inputs and acceptable economics. Agree who repairs wrong information and how quickly it must be corrected. Audit a sample of the generated collection as the data and template change, with stronger controls for consequential claims.

A small maintained catalogue can be the better investment when it helps customers decide and additional pages give them no further reason to choose the business. Hold or consolidate pages that duplicate existing answers, depend on unsupported facts or attract the wrong audience. Stop when maintenance and acceptance costs outrun plausible value.