Do Backlinks and Unlinked Media Mentions Help With AI Search?
By Jordan Baker, Founder of Fair Public Relations and former UK national journalist
There was a point in PR when a really good piece of media coverage was simply a really good piece of media coverage.
A journalist interviewed your founder.
A national newspaper quoted your expert.
A respected trade publication wrote about your company.
Your product appeared in a feature.
You showed the article to clients, shared it with your team and, quite reasonably, considered that a PR result.
Then digital PR became increasingly intertwined with SEO, and one question began appearing more and more frequently:
Did we get a backlink?
Backlinks can be extremely valuable.
I am certainly not anti-backlink. All other things being equal, I would rather have an excellent piece of online editorial coverage with a natural link to the client's website than the exact same article without one.
The link may send readers directly to the business. It can create a useful connection between two web pages. It may contribute to conventional search visibility. And if a customer is interested enough to learn more, it removes friction.
But I have never believed that an article suddenly becomes worthless because the journalist did not include a hyperlink.
And the growth of ChatGPT and AI-generated search makes that question even more interesting.
If an online newspaper article names your business, identifies your founder, explains their credentials and repeatedly associates them with a particular subject, there is now a public document containing all of that information.
There may be no link.
The information still exists.
The question businesses are now asking is whether those unlinked mentions can help AI systems understand, retrieve or recommend them.
The evidence is developing.
There are good reasons to believe independent public information matters to modern retrieval systems. There is interesting industry and academic research showing that the sources AI systems retrieve and cite matter substantially to the answers they produce.
But that is not the same thing as proving that every unlinked media mention acts as a universal AI ranking factor.
That distinction is the important part.
What is the difference between a backlink, brand mention and AI citation?
These terms are increasingly thrown together when they actually describe quite different things.
An editorial backlink is a hyperlink from an independently published article to another website, perhaps the company being discussed.
An unlinked brand mention is exactly what it sounds like. The article names the business but does not link to it.
An expert attribution may go further by connecting a named individual to their company, qualifications or area of expertise.
An AI source citation occurs when an AI system identifies or links to a source it has used in an answer.
And an AI recommendation happens when the system actually presents the business as an option relevant to what the user wants.
Those outcomes are not interchangeable.
Imagine a newspaper article says:
Sarah Green, founder of Green Finance, says first-time buyers frequently underestimate the cost of maintaining older properties.
In version one, the words Green Finance link to the company's website.
In version two, they do not.
The journalism itself may be identical.
The attribution is identical.
The expertise is identical.
The audience seeing the article is identical.
The link adds something useful.
Its absence does not delete everything else the article achieved.
Now imagine ChatGPT later cites the newspaper article when answering a mortgage question.
That is another distinct event.
And if ChatGPT recommends Green Finance to somebody seeking a mortgage adviser, that is different again.
A useful measurement framework needs to keep all of these things separate.
Why have backlinks mattered so much to SEO?
Links are fundamental to the architecture of the web.
They help people move from one resource to another and help search systems discover relationships between pages.
Not all links are treated equally.
A naturally included editorial link from a relevant article is different from buying hundreds of low-quality links solely to manipulate search rankings.
Google's current spam policies continue to treat link schemes intended primarily to manipulate rankings as spam. Examples include buying ranking links, automated link creation, low-quality directories and paid advertorial links that pass ranking credit. Google explicitly allows advertising and sponsored links where they are appropriately qualified, for example using `rel="nofollow"` or `rel="sponsored"`.
That distinction is important for PR.
A genuine editorial backlink is normally created because somebody publishing an article has decided there is a useful reason to link.
We are not selling the link.
We are not controlling the publisher.
We are earning an editorial opportunity, and the publisher decides how that article is presented.
Do nofollow links still have value?
Yes.
First, people can click them.
That alone seems strangely overlooked whenever marketers become overly fixated on technical link attributes.
If somebody reads an article, clicks through to your business and becomes a customer, the link was commercially useful regardless of how an SEO tool categorised it.
Second, the article itself still exists.
Third, search engines have become more sophisticated than the very old idea that links exist in a simple binary world where one passes value and another is invisible.
But I would still avoid promising a client that one particular link type will produce a specific ranking change.
The strength of the publication, context, topic, page, competition and wider website all matter.
Domain Authority is not the same thing as Google authority
This is worth stating again because digital PR reporting sometimes blurs the distinction.
Metrics such as Domain Authority or Domain Rating can be useful comparative indicators provided by third-party SEO companies.
They are not Google scores.
Google does not give a website a public Moz Domain Authority score and then rank it according to that number.
We track those metrics where they are useful, but they should be treated as external measurement tools, not a direct window into a search engine's algorithm.
The same principle applies to GEO.
A tool can create an "AI authority score".
That does not automatically mean ChatGPT, Google, Gemini or Claude uses that score internally.
What does an unlinked media mention provide even before we talk about AI?
Quite a lot.
Suppose a clinical psychologist is quoted in The Independent discussing relationships.
There is no backlink.
From a traditional PR perspective, the psychologist has still appeared in a respected publication.
Their name appears.
Their professional title appears.
Their expertise appears.
Their opinion appears.
A reader researching that person later may discover the article.
A prospective client can see that a journalist considered them credible enough to include.
Another journalist researching experts may encounter it.
The psychologist can legitimately point to the coverage as previous media experience.
None of those outcomes depends on a hyperlink.
This is why I think treating unlinked editorial coverage as a failure fundamentally misunderstands PR.
Have clients ever been disappointed because an article did not include a backlink?
Of course.
Particularly when SEO is one of the campaign objectives.
I understand it.
If a business has told us at the beginning that backlinks are particularly important, naturally it notices when a major article does not contain one.
My response is normally to explain the wider value of what they have received.
There is still independent coverage.
There is still brand visibility.
There is still credibility.
There is still public information associating the business or expert with the subject.
And potentially that wider body of public information becomes increasingly interesting in a world where AI systems are retrieving material from across the web.
I would simply stop short of telling them:
"Google definitely gives you a ranking point because your name appeared without a link."
We do not have evidence to state it that way.
Should a PR agency ask the journalist to add a backlink?
Sometimes, yes.
When I was working as a journalist, this was not an extraordinary request.
A PR might come back after publication and say:
"Would you be able to add the company's website?"
If the editor allowed it and the link made sense, fine.
Sometimes the answer was yes.
Sometimes the answer was no.
Sometimes nobody responded.
Publications have different policies, and those policies can change.
The same publisher may link in one type of article and not another. Some outlets use nofollow links. Some use followed links. Some are stricter about commercial websites. Editorial practices also change over time.
My general view is:
It does not normally hurt to ask politely where a natural link would genuinely help the reader.
But be prepared to accept no.
Do not badger a journalist who has just given you an excellent piece of coverage because you have decided the entire article is meaningless without a link.
That is a very quick way to damage a relationship over something the journalist may not even control.
A journalist does not always control the backlink
This is important for clients to understand.
The journalist writes the article.
An editor may change it.
A subeditor may add or remove links.
A publication may apply site-wide linking policies.
The CMS may automatically attach particular attributes.
A commercial team may govern links to certain categories of business.
So even if the journalist personally has no objection, they may not be able to give you precisely the link you want.
Good PR recognises that the publication belongs to the publisher.
Could ChatGPT use an article that does not link to your business?
Potentially, yes.
An AI search system can retrieve and process a web page because that page is relevant to the question being asked.
The page does not need to contain a hyperlink to every entity mentioned before the text on it becomes readable information.
OpenAI says public websites can potentially appear in ChatGPT Search when accessible to its search systems, and that cited sources may be shown alongside answers. Placement is not guaranteed.
So conceptually, an article saying:
"Jordan Smith, CEO of ABC Technology, believes..."
contains information about Jordan Smith and ABC Technology regardless of whether ABC Technology is hyperlinked.
What we cannot establish from that alone is:
whether ChatGPT will retrieve the page
whether it will recognise the company correctly
whether it will use that passage
whether it will cite the article
whether it will mention the business
whether it will recommend the business to somebody
Public availability creates the possibility of retrieval.
It does not prove retrieval occurred.
This is where GEO discussions often go too far
There is a tempting chain of reasoning:
AI systems retrieve information from the web.
My company was mentioned on the web.
Therefore the mention improves my ChatGPT ranking.
Step three does not automatically follow.
A page can exist and never be retrieved.
A system can retrieve it but choose a different passage.
It can use the information without displaying the company.
It can cite the page while recommending somebody else.
It can know the company from another source entirely.
That does not make unlinked mentions irrelevant.
It means we need to distinguish a plausible mechanism from demonstrated causation.
What does current AI-citation research tell us?
There is genuinely interesting evidence here.
Muck Rack's 2026 analysis examined more than 25 million links cited in responses from ChatGPT, Gemini and Claude across 17 industries. Under Muck Rack's classification, approximately 84% of cited links came from what it classifies as earned media, while journalism specifically represented about 27% and paid or advertorial content represented roughly 0.3%.
That sounds enormously encouraging for PR.
And it is interesting.
But the definition matters.
Muck Rack describes "earned media" as a broader category than journalism alone, including sources such as academic research, social content and corporate blogs.
So it would be wrong for a PR agency to turn that finding into:
"84% of ChatGPT recommendations are caused by newspaper PR."
That is not what the research shows.
It is evidence that non-paid public web sources constitute a substantial proportion of citations in Muck Rack's dataset.
That is worth knowing.
It does not create a guaranteed PR formula.
Academic research is even more useful for understanding the limitations
A July 2026 critical review examined 45 studies covering GEO and related generative-search research.
Its conclusion is very helpful because it pushes back against simplistic promises.
The authors argue that GEO is not one clean ranking task. Visibility can depend on several stages, including crawling, retrieval, reranking, context allocation, citation, model knowledge and user behaviour.
The review found evidence that already-retrieved content can influence whether and how it is used or cited, but it found no reviewed method demonstrating a stable, long-term, cross-platform causal effect on organic discoverability or downstream commercial behaviour.
That is a much more realistic description of where the industry currently stands.
New October research puts retrieval at the centre of the problem
Another preliminary 2026 study examined more than 158,000 generative answers and 468,000 citations across commercial queries.
It found that pages with titles and content semantically aligned with what people were asking tended to be cited more, but that retrieval explained far more variation than simple page characteristics within the dataset.
For ChatGPT specifically, the researchers reported that pages appearing as citations were pages its search process had first retrieved. They also observed that model knowledge of brands could influence which brands appeared in searches and answers.
This is early research, not official OpenAI ranking documentation.
But it reinforces a very important idea:
Being present somewhere on the web is not the same as being retrieved for a particular question.
An evidence table is more useful than a sales claim
| Claim | Evidence strength | What I would say |
|---|---|---|
| Natural editorial backlinks can provide referral traffic | Strong | Yes |
| Links are part of how the web and conventional search connect information | Strong | Yes |
| Google prohibits manipulative paid link schemes | Strong, official guidance | Yes |
| Unlinked press coverage can provide credibility and public brand information | Strong traditional PR judgement | Yes |
| AI systems can retrieve pages containing unlinked mentions | Technically plausible and observable | Yes, potentially |
| Earned/public web sources make up a substantial share of AI citations in some studies | Observed industry research | Yes, with methodology caveats |
| One unlinked press mention directly improves ChatGPT recommendations | Not established | Do not claim it |
| Ten editorial links guarantee AI recommendations | Not established | No |
| Domain Authority is a Google or ChatGPT ranking factor | False framing | No |
That is the level of distinction I think clients deserve.
Is a backlink therefore more valuable than an unlinked mention?
Often, yes, because it gives you something additional.
If I could choose between two otherwise identical articles and one includes a useful editorial backlink, I would choose the linked version.
The link creates a direct route to the client's site.
It may send referral traffic.
It can contribute to a stronger link profile.
It can also make the relationship between the source and business explicit.
But that does not mean I would turn down the unlinked version.
The real comparison is not:
link versus no link in isolation.
It is:
What is the entire editorial opportunity worth?
Are ten weak backlinks better than one excellent feature?
I would not assume so.
Imagine a cybersecurity expert has two options.
Option A: ten small generic websites publish short rewritten articles, each linking to the company.
Option B: one highly respected cybersecurity publication interviews the founder in depth about a subject directly related to their expertise, accurately names the company and discusses their thinking, but does not provide a backlink.
Which is better?
From a pure link-count perspective, Option A wins 10 to 0.
From a professional-credibility perspective, I might strongly prefer Option B.
If Option A's sites are genuinely useful and relevant, perhaps the links have value too.
But simply counting URLs tells us almost nothing about editorial quality.
The right answer depends on the campaign objective.
Would I still like the specialist feature to contain a backlink?
Absolutely.
This is not an argument against links.
It is an argument against reducing media quality to the presence of a link.
The ideal outcome is:
strong publication;
right audience;
relevant article;
accurate attribution;
substantive coverage;
and a natural useful backlink.
PR cannot guarantee every element.
One Less Thing is a useful example of where PR and backlinks overlap
Our campaign for cleaning franchise One Less Thing demonstrates why editorial links can be commercially interesting.
During the first 12 months, the campaign generated 38 published articles, of which 35 contained editorial backlinks.
During the same period, the newly launched website's Moz Domain Authority moved from 0 to 19.
That is useful evidence of the online footprint created during the campaign.
What it does not allow us to say is:
"Those 35 links caused every SEO change."
Nor could we claim:
"Those links caused ChatGPT to recommend the company."
Domain Authority is a third-party Moz metric.
Many other things can affect search performance.
The evidence should be described at the level it actually supports.
What about repeated mentions of an individual expert?
This is one of the areas I find particularly interesting.
A business name may appear once.
An expert can appear repeatedly.
Suppose a psychologist is quoted across multiple publications over six months.
Each article accurately identifies:
their name
professional title
organisation
consistent cluster of subjects such as mental health, relationships and human behaviour.
From a PR perspective, that creates a very clear media profile.
Fair saw this with Cambridge Therapy Centre, where Clinical Psychologist Christine Schneider generated more than 30 published media opportunities during a six-month campaign.
The traditional benefit is straightforward.
Journalists see that the expert has been used before.
Prospective clients see independent editorial visibility.
The person's professional positioning becomes clearer.
From an AI-search perspective, I would phrase the implication much more carefully:
there is now more public information repeatedly connecting that person with those subjects.
Whether one particular AI system retrieves those articles for one particular question remains a separate measurement question.
Accuracy becomes extremely important
If we are interested in public information being retrievable by AI systems, sloppy attribution becomes more problematic.
Imagine one publication describes somebody as:
"Dr Sarah Green, dermatologist at Skin Clinic."
Another calls them:
"Sarah Green, beauty expert."
Another says:
"Sarah Greene of Green Health."
Now we have inconsistent information.
Sometimes that is unavoidable.
Journalists write independently.
But from a PR perspective, we should provide accurate, consistent biographies, names, organisations and credentials wherever possible.
This is good media practice anyway.
GEO simply gives us another reason to care.
Should businesses buy backlinks for GEO?
I would be extremely careful.
There is a legitimate commercial world of advertising, sponsored content and paid media.
There is nothing inherently wrong with buying advertising when everyone understands what it is.
What I would not do is buy links designed primarily to manipulate search or AI systems and then pretend they are independent earned media.
Google explicitly identifies buying or selling ranking links and creating low-value content to manipulate links as link spam. Sponsored or advertising links should be appropriately qualified.
The same principle applies to artificial "authority".
If somebody creates fifty fake expert profiles, mass-produces generic articles and places them on low-quality sites solely to manufacture signals, the reputational risk is obvious even before we discuss algorithms.
## AI does not make fake expertise a better idea
If anything, I think it makes genuine expertise more important.
We are entering a world in which enormous quantities of content can be generated cheaply.
The scarce thing is not words.
It is credible first-hand knowledge.
Google's current people-first content guidance repeatedly emphasises original value, clear expertise and material genuinely useful to the intended audience, while warning against large volumes of automated content produced primarily for search manipulation.
That aligns very closely with how I think PR should work.
Find the person who actually knows something.
Then help them communicate it.
How should a business decide whether a press opportunity is worthwhile?
I would ask six questions.
First, is the publication genuinely relevant?
Second, does the article put the client in an appropriate context?
Third, is the attribution accurate?
Fourth, is there enough substance that the article would still be useful if Google disappeared tomorrow?
Fifth, does a natural backlink make sense and, if so, can we reasonably ask for one?
And sixth, does this support the wider thing the business wants to become known for?
That framework stops you making absurd decisions such as rejecting excellent national coverage because the publisher uses nofollow links.
Relevance matters more than collecting random domains
If a financial adviser wants to become recognised for retirement planning, an intelligent article in a respected personal-finance publication is extremely relevant.
A link from an unrelated cryptocurrency blog may be much less useful even if an SEO tool gives the website a respectable score.
Domain metrics can help assess a site.
They are not a substitute for reading it.
Ask:
Who publishes this?
Who reads it?
Why is my business mentioned?
Would I proudly send this article to a customer?
That remains an excellent sanity check.
How should you measure linked and unlinked coverage?
Do not force everything into one number.
For linked coverage, you can measure:
the publication
the context
the link
referral traffic
link profile development
resulting customer activity where measurable
For unlinked coverage, look at:
the publication
the prominence and accuracy of the mention
the subject association
potential branded search
repeated journalist use
wider visibility
Then separately monitor AI.
Is the article being cited?
Is the company being mentioned?
Is it being recommended for the right buyer questions?
Does AI understand the company accurately?
Are identifiable referrals arriving?
Are customers telling you AI influenced their decision?
Our separate guide, How Do You Measure AI Search Visibility?, goes much further into that measurement framework.
Should you chase a backlink after every article?
No.
Ask when it genuinely makes sense.
Do not make it an automatic ritual.
If a publication has an obvious policy of never linking, repeatedly asking achieves very little.
If the article already contains the client's full web address in plain text but the journalist accidentally did not hyperlink it, a polite query may be entirely reasonable.
If the story centres on a resource readers need to access, a link may be genuinely useful.
Use judgement.
PR is a human relationship.
What should journalists think when PRs ask for links?
Probably the same thing I did.
It depends.
A polite, sensible request is normal.
An aggressive message suggesting a journalist has somehow failed because they did not hand over an SEO asset is rather different.
Remember what happened first:
the journalist chose to include your client.
That has value.
So do backlinks and unlinked mentions help with AI search?
Backlinks have established conventional benefits.
Unlinked editorial mentions have established PR and credibility value.
Both create public information.
AI systems increasingly retrieve and synthesise information from the wider web, and current research shows that the source retrieval process is central to which pages get cited.
There is also significant descriptive evidence that earned and editorial sources appear frequently among AI citations.
What we still cannot responsibly say is that one backlink or one unlinked mention produces a guaranteed AI recommendation.
That would turn an emerging area of research into a sales promise.
I would much rather tell a client what we actually know.
Frequently asked questions
Do backlinks help a website appear in ChatGPT?
They may contribute to the wider online visibility and discoverability of a website, but there is no public formula saying a certain number of backlinks guarantees ChatGPT inclusion or recommendation.
Are unlinked brand mentions a Google ranking factor?
Google does not publicly document unlinked brand mentions as a standalone ranking factor in the way this question is commonly phrased. An unlinked article can still have clear credibility and brand value and creates public information about the business.
Can ChatGPT find a company mentioned in a newspaper without a link?
Potentially. If the article is accessible and retrieved, the text of the article can contain useful information about the company even without a hyperlink. That does not guarantee the article will be retrieved for a particular query.
Is a backlink better than an unlinked press mention?
All other things being equal, I prefer the linked version because it adds referral and potential SEO value. But excellent relevant coverage does not become worthless without one.
Does Domain Authority affect ChatGPT recommendations?
There is no public evidence that ChatGPT uses Moz Domain Authority as a ranking input. DA is a third-party SEO metric.
Do national newspapers count more than trade publications?
Not automatically. A respected specialist publication may be much more relevant to the subject, business and buyer.
How many press mentions does a business need for GEO?
There is no credible universal number.
Should you ask journalists to include a backlink?
You can ask politely where the link would be useful. Be prepared for the answer to be no and respect publication policy.
Can paid press mentions help SEO or GEO?
Paid exposure can have legitimate marketing value, but it should not be confused with independent earned editorial coverage. Paid links intended to manipulate Google rankings can breach Google's spam policies if not appropriately qualified.
How can I track the value of unlinked coverage?
Look at the quality and relevance of the article, brand and expert positioning, branded search, subsequent media interest, customer attribution and separately observed AI mentions or recommendations.
Looking for genuine media coverage, not artificial links?
Fair Public Relations develops credible media angles, expert commentary, stories and news that journalists can choose to use.
Sometimes that coverage includes an editorial backlink.
Sometimes it does not.
We will always recognise the value of a useful natural link, particularly where digital visibility is part of the client's goal.
But our job is not to manufacture links.
It is to build legitimate editorial visibility around businesses and people who have something worthwhile to contribute.
If that is what you need, our Expert Commentary PR and Standard PR services explain how we work.
About the author
Jordan Baker is the founder of Fair Public Relations and a former UK national journalist with more than a decade of experience in journalism, PR and media relations. He writes about earned media, backlinks, media relations, expert positioning, SEO, GEO and how independent editorial coverage contributes to a business's wider public profile.
View Jordan Baker’s media expert profile
