Does AI written content rank?
It can, and the distinction that decides it is not how the text was produced. Search engines have never been able to detect authorship reliably and do not primarily try. What they can measure is whether a page satisfies the person who opened it, and whether anything on it exists that was not already on twenty other pages.
So the honest answer is: AI assisted content ranks when a person has added judgement, evidence or experience to it. Content generated at volume and published unread does not, and it does something worse than failing on its own, which we get to below.
Why publishing at volume backfires
Publishing a hundred unreviewed pages is the fastest way we know to build a site that ranks for nothing and quietly damages the pages that used to work. Quality is assessed at the site level as well as the page level, so a large body of thin material dilutes the signal of the handful of pages that were genuinely good.
There is also a crawl cost. Attention paid to a hundred pages nobody wants is attention not paid to the six that convert. Sites in this state usually do not fall off a cliff. They flatten, slowly, and the cause is hard to see precisely because nothing dramatic happened.
Where AI genuinely helps
Research and structure, which is where the leverage was always going to be. Clustering thousands of real queries into topics a human can act on is work that used to take days and now takes an afternoon. Drafting an outline in a minute rather than an hour is real time back. Finding the gap between what you have published and what people actually ask is a genuinely hard task that models are good at.
On our own tool pages we use the same split. The blog outline generator produces the skeleton, the meta description generator drafts the summary, and a person writes the parts that carry the claims.
Where a human still has to decide
Four things, and none of them compress. What is true. What your business can honestly claim. Which of the fifty candidate topics is worth the effort at all. And where the draft is confidently, fluently wrong, which is the failure mode that costs most, because fluent wrongness reads exactly like competence.
This is the part that determines whether any of it ranks, and it is also the part that gets cut first when the goal becomes output volume.
Entities matter more than keywords now
AI assistants answer from a model of what your business is, not from a list of ten blue links. That shifts the work. Consistent naming across every profile, clean structured data, an address and phone number that match everywhere they appear, and corroborating mentions on sites you do not control now do the job that keyword density used to.
Practically, that means the boring pass matters: one spelling of your business name, one address format, one phone number, and markup that says the same thing your page says. Our free schema tools cover the common cases without a plugin, including LocalBusiness, Article, FAQ and Product markup.
How to be quotable by AI answers
Extractive systems lift passages, not pages. A paragraph that only makes sense after reading the three above it is a paragraph that cannot be quoted, however good the article is. Writing for citation is mostly a structural discipline rather than a stylistic one.
- Answer the question in the first sentence under the heading, then explain. A reader who leaves after one line should still have the answer.
- Make headings the questions people actually type, not clever labels.
- Keep each passage self contained, so it survives being quoted without its neighbours.
- Use lists where the content genuinely is a set of steps or criteria, because lists are what extractors lift most reliably.
- Put a visible date on anything time sensitive. A date that exists only inside markup is invisible to a text extractor.
- State facts as facts, with numbers where you have them and without inventing them where you do not. Unverifiable claims are the ones that get dropped.
A practical loop that fits a small team
None of this needs a large content operation. It needs a repeatable weekly pass that a single person can run without dread.
Find the questions your customers ask on the phone and write them down verbatim, because that is your keyword research and it costs nothing. Check how a page currently presents itself in results with the SERP preview tool before you rewrite anything. Run the site through the website scorecard to find the technical faults worth fixing first, and use the robots.txt generator if crawling rules are part of the problem. Then publish one genuinely useful page rather than five padded ones.
- Collect the real questions from calls, WhatsApp and email each week.
- Pick the one with the clearest commercial intent.
- Draft the structure with AI, write the claims yourself.
- Check the result in the SERP preview, add the right schema, publish.
- Come back in a month and improve the page rather than writing another one.
What we would not do
We would not spin one article into fourteen city variants with the place name swapped, we would not publish a page we had not read, and we would not add markup describing content that is not on the page. All three work briefly and all three are the sort of thing that gets corrected in a later update, usually across the whole site at once.
If you want the longer version of how we approach this, it is our AI and SEO service, and the short version is that the tooling changed and the standard did not.
Common questions
Will Google penalise content written with AI?
Not for being written with AI. Google's stated position is that it rewards helpful, reliable content regardless of how it was produced, and acts against content produced at scale primarily to manipulate rankings. The practical risk is not the tool, it is publishing unreviewed material in volume.
What is GEO, and is it different from SEO?
GEO refers to optimising for generative answer engines such as AI Overviews, ChatGPT and Perplexity. It overlaps heavily with good SEO, but it puts more weight on passage level clarity, self contained answers, consistent entity information and structured data, because those systems quote passages rather than ranking pages.
Do I need an llms.txt file?
It is optional and currently ignored by Google Search. It costs little to publish, but it should be well down the list behind having crawlable, well structured pages that state their claims clearly, which is what every one of these systems actually reads.
How long does SEO take to show results?
For a small local site, expect months rather than weeks for competitive terms, though technical fixes and correcting inconsistent business information can move things sooner. Anyone promising a specific position by a specific date is describing something they cannot control.