What you can do with Antidote
Human approval means someone in your business checks AI drafts before deciding they are ready. You get a chance to correct facts, add experience and keep your voice. The studies below examine rankings, traffic and audience responses in different settings; they cannot tell you a required share of human input.
80.5%
In a Semrush analysis of 42,000 blog posts, the page in position 1 had an 80.5% probability of being human-written and a 10% probability of being AI-generated, making human-written content about 8 times more likely to rank first.
In a Semrush analysis of 42,000 blog posts, the page in position 1 had an 80.5% probability of being human-written and a 10% probability of being AI-generated, making human-written content about 8 times more likely to rank first.
86.5%
Most ranking pages use some AI, but purely AI pages rarely reach #1.
Ahrefs analysed the top 20 results for 100,000 keywords (600,000 pages): 86.5% contained at least some AI-generated content, and purely AI-generated content rarely reached position 1.
82.2%
Pages that are mostly human hold 82.2% of top-3 rankings.
In an Ahrefs study of 331,000 pages, pages with under 50% AI content accounted for 82.2% of top-3 rankings.
283 visits
By month five, human-written posts averaged 283 visits each; AI-only posts averaged 52.
NP Digital tracked traffic per post: by month five, human-written articles averaged 283 visits per post against 52 for AI-generated articles, and the gap widened over time.
less natural and less useful
The same ad, labelled AI-made, was rated less natural and useful, and people engaged less.
In experiments by the Nuremberg Institute for Market Decisions, identical ads labelled as AI-generated were rated less natural and less useful than the same ads labelled human-made, and willingness to research or buy the product fell.
46%
People are twice as likely to choose a business that replies to all its reviews.
BrightLocal's research finds consumers are twice as likely to use a business that responds to all of its reviews than one that responds to none, and whether a business responds matters to 46% of local consumers.
These snapshots help a reviewer decide what to check: AI recommendations separately from Maps rank, unanswered reviews and the accuracy of profile posts. They do not test whether human approval changes rankings, reply rates or bookings.
13.8%
of businesses ChatGPT named were in Google Maps’ top three for the same search.
Does ChatGPT recommend the businesses Google Maps ranks first?Watch the short filmThe research below includes search ranking studies, a traffic comparison and an experiment about advertising labels. Each asks a different question. A link between writing style and rankings cannot prove an editor’s change caused the result. Read the method and audience before using a study to decide your own publishing policy.
Ask whether the piece answers a real customer question, whether the facts have evidence and whether your business can stand behind the wording. Add useful experience missing from the brief. For a public reply, also check privacy and whether joining the conversation helps. Remove claims you cannot defend before publishing.
Anything you ask for in the app waits for your approval. For regular work, choose with the team whether each draft waits or we publish on your behalf. Review replies follow a separate agreed policy. Those settings decide when someone checks a piece; they do not guarantee rankings or a particular marketing result.
No. They use different methods and do not isolate the effect of a final edit on a particular page.
The studies do not give a required percentage. Focus on accuracy, useful experience and a responsible publishing decision.
No. The Ahrefs findings include AI use among ranking pages. AI drafting and human review can be part of the same process.
See the app with examples from your industry, and what we would run for you.