A page can be indexed, fast, and free of duplicate URLs, and still fail to rank or to be cited. Thin content does not answer the query. Outdated content answers it with a fact that is no longer true. Unhelpful content was written to match a keyword, not to help the person who searched. Those are SEO issues when Google will not keep the URL in a competitive result, and GEO issues when ChatGPT, Perplexity, Gemini, or a Google AI Overview quotes the stale sentence or skips the brand because there is nothing precise to extract.
This guide follows common SEO and GEO issues that hurt rankings. Use the indexing and crawl guide when the URL is not in the index, duplicate content and keyword cannibalization when several URLs fight for one query, the schema markup guide when the JSON-LD disagrees with the paragraph, and the Core Web Vitals guide when field data for speed or stability is poor. Use this one when the page is in the index and still does not deserve the citation.
Three faults that look alike in a ranking report
| Fault | What you see on the page | SEO effect | GEO effect |
|---|---|---|---|
| Thin | A few sentences, no steps, no example, no local detail | Rarely holds a competitive query | Nothing specific enough to quote |
| Outdated | Old prices, old years, discontinued names, a “current” fee that changed | Freshness-sensitive queries drop; others may linger | The old number is the sentence that gets cited |
| Unhelpful | Unique text that only repeats the keyword | Demoted against pages that actually answer | A clearer source is retrieved instead |
Word count is not the test. A short definition that states the term, who it is for, and how it differs from the neighbouring term can rank and be cited. A 2,000-word page that restates the same claim is still thin for the intent. Judge the page by whether a reader can act on it without opening a second result.
Thin content is an SEO issue, not a short page
Search systems reward pages that satisfy the query. A service URL that says the company “provides professional solutions” and then lists the same five bullets as the homepage does not. Neither does a district page that swaps “Central” for “Tsim Sha Tsui” and leaves every other sentence unchanged. Those URLs may be crawled. They seldom earn a stable SEO ranking, because a more specific page already answers the searcher.
On Hong Kong sites the thin page is often one language. The English guide explains the steps, the fee range in HKD, and the documents a client needs. The Traditional Chinese URL is three translated paragraphs and a contact form. Queries asked in Chinese then have no page that deserves to rank, and the English URL is a weak match for that language. The reverse is just as common. Treat each language as its own page, not as a shorter copy of the other.
Thin content also shows up as doorway sets: one URL per industry synonym, each with the same outline. That overlaps keyword cannibalization. The difference is intent. Cannibalization is several real pages competing. Thin content is pages that should not have been separate URLs.
Outdated content and content decay
Outdated content is a page whose facts have moved on:
- A price, a statutory fee, or a tax rate from a previous year
- A product or scheme name the organisation no longer uses
- Statistics with no source year, or a year that is now the oldest result on the query
- A title that still says “2024 guide” after the body was partly rewritten
- Hours, addresses, or districts that changed and were fixed in the footer but not in the article
Content decay is the ranking version of the same problem. The URL once matched the query better than anything else. Competitors later published a clearer comparison, a newer figure, or a worked example. Positions slip even though the technical audit is clean. There is no noindex, no slow template, and no duplicate title. The page is simply no longer the best answer.
Freshness is query-specific. A definition of a stable term does not need a monthly rewrite. A page about a fee, a regulation, or a product lineup does. Changing dateModified on every deploy, while the paragraph still states the old fee, makes the page look maintained and leaves the wrong fact in place. The schema guide covers that date mismatch. This guide covers the sentence itself.
Unhelpful content
Unhelpful content can be long, original, and grammatically fine. It was produced to occupy a keyword: an article that defines the term in the first sentence and then repeats it, a roundup that ranks providers without saying what was compared, or a page generated from an outline with no example a practitioner would recognise.
Google’s ranking systems have folded “helpful content” ideas into core ranking. Pages created mainly to match queries, without adding information someone could not already get from the rest of the results, tend to underperform. That is an SEO ranking problem even when the HTML is valid.
It is also distinct from a manual spam action. Most of these pages are never penalised as a site. They are simply outranked. A sitewide pattern — dozens of near-empty service URLs, or a blog that publishes the same outline every week — can still drag the sections around them, because internal links and crawl attention point at pages you would not cite yourself.
Why these pages are a GEO issue
Generative engines quote a passage. If the only number on the page is a 2023 price, that price is what a model can store. The citation looks specific and is wrong. Updating the English page and leaving the Chinese page untouched creates a second source that disagrees with the first. Answer engines do not always prefer the language of the prompt’s newest edit. They prefer a sentence they can lift.
Thin pages fail earlier. A model that needs a definition, a step list, or a comparison has nothing to extract, so it cites the competitor who published the table. Content formats AI platforms cite shows those shapes. E-E-A-T for AI visibility covers the byline, the date, and the first-hand detail that make a passage worth trusting. None of that helps a page whose only original sentence is a slogan.
Unhelpful content has a further cost. Pages that stuff the same keyword into every heading are easy for a retriever to mistake for the canonical explanation, then discard once the passage does not answer the prompt. The brand is retrieved and not cited. Teams read that as an AI visibility problem. It is a content-quality problem.
A practical order of fixes
- Export the queries that fell, and the URLs with impressions and a weak click-through rate. Separate technical faults first. If the URL is excluded from the index, start with the crawl guide. If three URLs target one query, start with cannibalization.
- Label each remaining URL as thin, outdated, unhelpful, or worth keeping. A page can be two of these. A fee table from 2023 that also lacks steps is both outdated and thin.
- Keep one URL per intent. Merge the doorway and synonym pages into the page that should rank, redirect the extras, and point internal links at the survivor. Do not leave the merged URL as an orphan.
- Rewrite the opening paragraph so it answers the query in two or three sentences: the definition, the current figure, and who it applies to. Put the year next to any number that can go stale.
- Add one element a competitor’s page already has and yours lacks: a table, a numbered procedure, or a Hong Kong example with HKD, a district, or a bilingual query. That is the difference between a rewrite and a longer version of the same thin page.
- Update English and Traditional Chinese in the same release. If a fact changes, change it in both. If one language cannot be reviewed yet, do not leave the old figure live on that URL.
- Set the visible updated date, and
dateModifiedin the structured data, only when the facts changed. Align the title year with the body. - Noindex or remove URLs that exist only to hold a keyword and have no links, no conversions, and no answer you would put your name on. Keep a page that still earns its place, even if it is shorter than the blog average.
Then ask the prompts customers actually use and read which sentence was quoted. If the quoted sentence is the one you just corrected, the GEO issue was the stale fact. If the citation still points elsewhere, the remaining gap is entity and coverage, which the broader SEO and GEO issues guide is meant to audit.
Questions teams ask
Does a low word count cause an SEO ranking problem?
No. A short page that answers the query can rank. A long page that never answers it can fail. Thin content means the page is incomplete for the intent: missing steps, missing evidence, or a location page that only changes the place name. Add the missing answer, or combine the URL with a page that already has it. Do not pad the word count.
Will deleting outdated content improve AI rankings?
Deleting a page that still matches a query removes it from both Google and from any crawler that might have cited it. Update the facts when the URL should remain the source. Redirect or merge when a better URL already exists. Remove the page when it has no demand and no answer worth keeping. After the change, confirm the old price or year is gone from both languages, or an AI answer can keep quoting the copy you left behind.
How often should a bilingual page be updated?
When the fact changes, not on a calendar. Fee, regulation, product, and address pages need a review whenever that fact changes, and both languages need the same review. Evergreen definitions need a review when the surrounding results have become more specific than yours, which is content decay, not a scheduled rewrite. A date in the title that you cannot defend is a reason to edit now.
Is AI-written content an SEO or GEO issue by itself?
The issue is whether the page is accurate, specific, and attributable. An assisted draft that a practitioner checks against current fees, names the author, and adds an example can be useful. An unedited draft that repeats a keyword and invents a figure is unhelpful content, and it is a GEO risk because the invented figure is easy to quote. The byline and the sources matter more than the tool that produced the first draft.
What to review this month
List the twenty URLs that lost the most clicks, plus any page whose title still carries an old year. Mark each one thin, outdated, unhelpful, or sound. For the outdated set, write down the figure that is wrong and correct it in English and Traditional Chinese together. For the thin set, either add the missing answer or fold the URL into a stronger page and redirect it. Re-fetch both languages and confirm the old fee, the old year, and the swapped-district boilerplate are gone.
Thin, outdated, and unhelpful content is an SEO issue with a GEO consequence: search stops ranking the URL, and answer engines keep the sentence you no longer mean. Treat the paragraph as part of the same audit as the canonical, the schema, and the other language.