For Local SEO, this matters because service-by-town and brand-by-town page patterns are extremely easy to scale. The fact that a tool can generate 100 pages does not mean 100 useful pages exist.
Key takeaway
Use AI to organise, draft, edit and check genuine information. Do not use it to manufacture customers, jobs, locations, reviews, expertise or “local” facts that never existed.
What Google currently says about AI-generated content
GOOGLE CONFIRMS
Google's current guidance says generative AI can be useful for research and adding structure to original content. Google does not say that all AI-generated content is banned.
Google's focus is on accuracy, quality, relevance and compliance with Search Essentials and spam policies.
Its guidance specifically warns that generating many pages without adding value for users may violate the scaled content abuse policy.
That is an important distinction. The creation tool is not the sole issue. The purpose and value of the published content are the issue.
GOOGLE CONFIRMS
Google's people-first guidance asks creators to focus on content made primarily to help people rather than content made primarily to attract search visits. Google's current guidance also says that automation used mainly to manipulate rankings can violate spam policies.
UNKNOWN / NOT ESTABLISHED
Google does not publish an “acceptable AI percentage”, a rule saying a page must be a certain percentage human-written, or a universal AI-detection threshold. Do not invent one.
Scaled content abuse is about scale plus purpose plus low value
GOOGLE CONFIRMS
Google defines scaled content abuse as generating many pages for the primary purpose of manipulating search rankings rather than helping users. Its examples include using generative AI to create many pages without adding value, but also scraping, transformations, stitching content and other methods.
The policy is deliberately method-neutral. A human can create scaled low-value content. An AI tool can help create excellent content. Automation can be useful or abusive depending on what is produced and why.
Why Local SEO is particularly exposed
Local keyword lists naturally produce combinatorial patterns:
- service x town;
- service x neighbourhood;
- manufacturer x town;
- product x town;
- “near me” variations;
- multiple pages with the same structure and swapped place names.
That makes Local SEO vulnerable to a dangerous shortcut: treat every keyword combination as a page.
Examples of weak scaling include:
- 50 pages where only the town name changes;
- a generated paragraph about local landmarks that has nothing to do with the service;
- invented statements that the business “regularly works throughout” places it has never served;
- fake projects used to make a location page look unique;
- fabricated testimonials;
- made-up business history;
- dozens of manufacturer pages created only because the brand names have search volume.
KEYWORD COMBINATIONS ARE NOT AUTOMATIC URLS
Doorway-page policy is a related but separate topic. A full analysis of location-page doorway risk belongs in the dedicated guide.
Good uses of AI in a local content workflow
AI can reduce repetitive editorial work without replacing first-party evidence.
Useful examples include:
- brainstorming possible customer questions from a supplied service brief;
- turning genuine job notes into a first draft;
- reorganising a long draft into a clearer structure;
- summarising supplied product specifications;
- checking consistency of names, URLs and terminology;
- suggesting internal-link opportunities between already-approved owner pages;
- comparing two drafts for duplication or cannibalisation risk;
- editing for clarity and UK English;
- identifying claims that need a source;
- producing alternate heading options from the same verified intent.
Google's own current generative-AI guidance explicitly recognises research and content structuring as potentially useful applications.
The human/editorial process still has to verify factual claims.
Bad uses of AI
Do not use AI to invent evidence.
That includes:
- jobs that never happened;
- customers who do not exist;
- testimonials or reviews;
- locations the business does not serve;
- addresses;
- business history;
- statistics;
- qualifications or expertise;
- project outcomes;
- partnerships;
- backlinks;
- photographs described as genuine when they are not;
- “local knowledge” copied or hallucinated from unrelated sources.
The more a page depends on first-hand experience, the more important it is that the source material comes from the real business rather than the language model.
Green, amber and red examples
GREEN - useful assistance
A furniture assembler supplies real job notes, product model, location, build time and approved photographs.
AI helps organise the facts into a draft case study. A human checks every statement before publication.
GREEN - editorial consistency
A site has 60 approved pages. AI helps identify repeated anchors, duplicate headings and places where the same paragraph has been reused. The editor decides what to change.
AMBER - templated draft requiring substantial evidence
AI creates a first draft for a proposed service-area page using only the town and service name. Before publication, the business must add genuine service evidence, useful local context, distinct customer information and a clear reason the page deserves to exist. If it cannot, the page should not be published.
RED - fabricated local evidence
AI invents three “recent customer projects” in towns the business has never served and generates testimonials to make the pages look unique. This is misleading regardless of whether the text reads well.
RED - mass keyword multiplication
A spreadsheet produces 20 services x 30 towns and an AI tool generates 600 pages because each combination exists as a keyword. The pages are mostly interchangeable and exist mainly to capture search traffic. That is exactly the kind of scaled strategy that should trigger a policy and quality review.
Organise genuine notes, edit supplied facts and check structure.
A template draft has incomplete local facts and needs human input.
Invented jobs, testimonials, local facts or mass town-swapped pages.
Use an evidence-first publishing workflow
A reliable workflow starts before the prompt.
1. Define the page intent
What single user question should this URL own?
2. Gather genuine first-party facts
Collect real services, locations, job notes, photographs, customer questions, product details and measurements.
3. Source platform/policy claims
If the page discusses Google policy, use current official documentation. Do not ask AI to recall a policy from memory and publish the answer unchecked.
4. Draft with or without AI
AI may assist with structure or language, but the evidence remains external to the model.
5. Human fact-check
Check names, dates, products, locations, links, quotes, policies and numbers.
6. Add original evidence
Use real photographs, examples, project notes, data or observations where available.
7. Cannibalisation check
Make sure the page owns a distinct question and is not duplicating an existing URL.
8. Internal-link check
Link to the authoritative owner pages rather than repeating their full content.
9. Editorial approval
A named human/editorial process takes responsibility for the final claims.
10. Publish and measure
Track whether the page is useful to users and whether it contributes to real visibility and business outcomes.
The “could this page exist without real evidence?” test
Before publishing a local page, remove the town, service or brand name mentally.
If the page now reads like a generic template that could be used for fifty other locations with almost no change, ask whether there is enough distinct value to justify the URL.
A page may still be legitimate without a completed project in that exact town, but it needs a real reason to exist: distinct demand, genuine service coverage, useful local information, clear service relevance and content that helps a customer make a decision.
If the only unique element is a swapped place name, the page probably needs more substance or may not deserve to exist.
AI disclosure and authorship
Google's current people-first guidance suggests that explaining how content was created can be useful where readers would reasonably ask “how was this made?”. That does not mean every AI-assisted sentence needs a warning label.
The more important principle is accountability.
Do not invent an author biography or claim hands-on experience the author does not have. If AI helped draft or organise material, the human editorial process remains responsible for:
- accuracy;
- sourcing;
- permissions;
- privacy;
- final claims;
- whether first-hand evidence is genuinely first-hand.
Real Project: McKnight's Flat Pack Assembly
The Flat Pack project shows why an evidence-first approach matters.
GOOD AI USE could include:
- organising genuine Dunelm project notes into a case-study structure;
- summarising supplied product specifications;
- editing copy for clarity;
- checking internal links across approved pages;
- comparing location-page drafts for repetition.
BAD AI USE would include:
- inventing flat-pack jobs in towns where no verified work exists;
- creating fake customer reviews;
- making up customer names;
- claiming fabricated build times;
- creating dozens of town-swapped pages purely because keyword combinations exist.
At this stage, no specific historic AI workflow for McKnight's Flat Pack Assembly is supplied as a verified case-study record. Therefore these examples should be treated as acceptable/unacceptable use cases, not as claims that the project has already used each method.
Practical publishing checklist
Before any AI-assisted Local SEO page goes live, confirm:
- one distinct user intent;
- an approved owner URL;
- genuine business facts;
- no invented jobs, reviews, customers or locations;
- current official sources for platform/policy claims;
- useful first-party evidence where relevant;
- no mass town/service substitution;
- no substantial duplication with an existing page;
- internal links to related owner pages;
- human fact-check complete;
- privacy and permissions checked;
- editorial approval recorded.
Next step
Use the checklist before publishing any AI-assisted page. For any proposed location page, also run the idea through the Location Page Qualification tool before a URL is created.