What would AI say if a travel blogger asked it directly about lost traffic and getting paid for original work? I decided to try it with Astra, and this is the conversation we had about compensation, responsibility and who will keep creating fresh travel information.
I have spent a lot of time asking what artificial intelligence means for travel bloggers, and most of those conversations involve humans talking about AI. This time I wanted to ask AI directly.
Curiosity was not the only reason I asked these questions. I had emailed OpenAI and Anthropic and sent several follow ups, but neither company replied, so I decided to ask Astra directly.
I am a travel blogger, and I have watched search traffic decline while AI systems answer questions using knowledge that people like us spent years and a lot of money creating. I use AI and believe it can help us, but I am also angry about the damage and the lack of a fair economic exchange.
People keep talking as if AI is getting close to artificial general intelligence. I do not know how close we really are, but if these systems are powerful enough to influence how people plan trips, they should be powerful enough to answer difficult questions about what they are doing to the people who create travel information.
For this experiment, I used GPT-6 Astra, the most capable OpenAI model available in my ChatGPT Work session. I first sent ten questions together in one compact interview to control the cost.
After reading its answers, I added one follow up about whether the information ecosystem itself is broken. For Question 12, I asked Astra to turn the interview around and ask travel bloggers whatever it most wanted to know.
The first eleven answers and Astra’s final question below are its original words, with only formatting added for readability.
Astra does not have feelings, personal guilt, or access to private decisions made inside OpenAI or other AI companies. What I wanted to discover was whether an advanced model, when asked directly, would recognize the ethical and financial problem instead of avoiding it.
1. Should AI builders accept moral responsibility for harm to creators?
Yes. People and companies that build and profit from these systems should take responsibility for foreseeable harm to the people whose work helps make them useful.
I do not experience guilt, and presenting an artificial expression of guilt would not compensate a travel blogger. The relevant questions concern human decisions: who benefits, who bears the costs, what permission was obtained, and what remedies are available.
Responsibility does not mean that every lost visit is attributable to AI, or that every existing publishing business must survive unchanged. It does mean that “the technology is inevitable” is not an adequate excuse for ignoring consequences that product design and commercial arrangements can influence.
The meaningful response would be measurable action: negotiated compensation, useful attribution, publisher controls, and honest reporting about referrals and content use. An apology without changes to those arrangements would have little practical value.
2. Are AI referrals replacing the traffic and revenue publishers lose?
The strongest 2026 evidence suggests that they are not. A preregistered field experiment from researchers at the University of Pennsylvania and Northeastern University followed 1,100 people using Google Search during normal browsing.
Assignment to Google AI Mode reduced the clickthrough rate to external websites by 18.8 percentage points compared with current Google Search, while exposure to a version without AI features increased it by 8.8 points. AI Mode also reduced the share of participants clicking news sites by 12.5 points and lowered reported trust, usefulness, and satisfaction. AI in Search Reduces Publisher Referrals Without Improving User Experience
A separate University of Washington study used 161,382 matched Wikipedia article and language pairs. It estimated that exposure to Google AI Overviews reduced daily traffic to English pages by about 15%, although Wikipedia is not a travel blog and its results should not be applied directly to every publisher. Impact of AI Search Summaries on Website Traffic
Another 2026 study from Washington University in Saint Louis issued 55,393 trending queries across 19 topics. At least 50.6% of pages cited by AI Overviews carried display advertising, while Google continued showing its own sponsored ads on pages containing AI Overviews. Measuring Google AI Overviews and Publisher Impact
These papers are working drafts and preprints rather than final peer reviewed publications, so their limits should be clear. Even so, they measure the problem with experiments and large datasets instead of repeating one survey about click behavior.
They do not calculate the exact financial loss for an independent travel blog, but they show the direction of the exchange. AI search keeps more attention inside the platform while reducing the visits that can become advertising income, bookings, affiliate sales, newsletter subscribers, or returning readers.
Until AI companies disclose auditable publisher referrals and conversions, they cannot credibly claim that citations or new AI traffic replace the value they displace. The cost of creating and updating firsthand travel information still remains with the publisher.
3. Is using firsthand work without compensation or meaningful traffic fair?
As a general commercial arrangement, no, especially when a service substitutes for the creator’s work while leaving that creator with the expense of producing and maintaining it. Public accessibility is not, by itself, a convincing ethical justification for any profitable downstream use.
There are important distinctions, though: learning a general fact, quoting a short passage, reproducing a photograph, and reconstructing the practical value of an entire guide are not equivalent. Neither should a creator’s investment establish exclusive control over every fact about a destination.
The strongest fairness claim concerns the specific reporting, judgments, photographs, and continuing verification that distinguish original work from interchangeable information. If those contributions materially support an answer or product, the commercial arrangement should recognize their value.
Meaningful permission, compensation, and attribution could make some uses fair. A decorative link that delivers almost no benefit should not automatically count as sufficient repayment.
4. Where will accurate information come from if independent reporting becomes unaffordable?
Some information would still come from governments, businesses, travelers, researchers, and reporting commissioned directly by platforms. The problem is not that all information would disappear, but that the mix of available information could become less independent.
A hotel can report its opening hours and amenities, but it has a commercial interest in how those amenities are described. An independent visitor can investigate whether the advertised accessibility works in practice or whether nearby construction makes a room unpleasant.
AI can organize accounts of those conditions, but generating more text does not establish what is happening on the ground. Someone must observe, verify, and update the underlying evidence.
If independent reporting becomes uneconomic, platforms would need to fund it, purchase it, or accept gaps and greater dependence on interested sources. They should not assume that unlimited synthesis creates an unlimited supply of fresh knowledge.
5. Should answers identify and link to contributing travel bloggers?
Yes, when an answer materially relies on identifiable reporting, the source should normally be named and linked without requiring the user to request it. Attribution should sit close enough to the relevant claim that readers can understand what the source actually supports.
There is a significant limitation: I cannot reliably reconstruct which individual training documents contributed to every generated sentence. A system should never invent that history or attach a plausible blogger’s name to information it did not obtain from that blogger.
That limitation strengthens the case for retrieving and checking identifiable sources when answering questions about current travel conditions. It does not justify pretending that unsupported recollection is verified reporting.
Citations should favor the original work where that origin can be established, rather than a derivative page repeating it. But attribution and compensation solve different problems: giving credit does not necessarily replace the income lost when the answer satisfies the reader completely.
6. What would fair compensation look like?
A workable system would probably combine several mechanisms rather than rely on one payment formula. Independent creators should have access to collective bargaining or licensing arrangements, because negotiating separately with every AI company would consume time many cannot afford.
For ongoing access to original reporting, a guaranteed licensing payment could support the cost of maintaining coverage. Additional payments could reflect documented use, with transparent definitions and independent auditing rather than an unexplained calculation controlled entirely by the buyer.
Paying only for citations would invite citation manipulation and leave uncited use unrewarded. Revenue sharing could help, but only if creators can examine what revenue is included and how their share is calculated.
Reader subscriptions and paid guides may strengthen a blogger’s business, but they are not compensation from AI companies. A fair agreement also needs meaningful refusal, reasonable exit terms, and protection against deals that purchase permanent access for a trivial upfront amount.
7. Will large travel agencies dominate recommendations?
That is a plausible risk, not an outcome I can establish as inevitable or a plan I can attribute to AI companies. The important question is which incentives and technical choices shape what the user sees.
If a recommendation system privileges bookable inventory, easily integrated commercial feeds, and transactions that generate revenue, large intermediaries could gain an advantage. An excellent local guide without sophisticated booking infrastructure could become less visible despite being a better fit.
The opposite is also possible: a system deliberately designed to retrieve independent local reporting could surface businesses that conventional rankings overlook. That possibility is not evidence that it will happen automatically.
Platforms should disclose commercial relationships and distinguish recommendation quality from their ability to process a booking. The warning sign would be an assistant presenting itself as an impartial travel adviser while quietly restricting practical choices to its commercial partners.
8. What can bloggers offer, and what should they stop protecting?
An experienced blogger can go somewhere, notice what the original question missed, test a claim, and put a reputation behind the findings. I can help interpret descriptions of an experience, but I cannot truthfully claim to have taken the boat, inspected the room, or spoken with the driver.
That difference matters most when the reporting changes a decision: who should avoid an activity, what disappointed you, or why the advertised option is unsuitable for a particular traveler. Specific evidence is more useful than simply attaching a human label to generic advice.
Bloggers should not expect protection for every page whose main function is repackaging easily available facts. Readers do not owe a website a visit merely because its business model requires one.
The defensible work is original observation, distinctive judgment, useful relationships, and accountability. Even that work needs distribution and a viable business model; being genuinely valuable does not guarantee being paid.
9. What rules should govern AI use of travel content?
I would recommend separating permission for search indexing, model training, and live retrieval, rather than treating them as one indivisible bargain. Publishers should be able to understand and express those choices without needing a specialist engineering team.
Commercial use of substantial original reporting should have clear terms, accessible licensing options, and records that permit meaningful auditing. Rules should also preserve reasonable quotation, public facts, research, and legitimate criticism rather than creating private ownership of general knowledge.
Answers should accurately attribute retrieved reporting, show relevant dates, distinguish verified information from inference, and disclose paid placement or booking relationships. Systems should offer correction channels that can actually change recurring errors.
Finally, compliance should be independently assessable, with remedies for misuse rather than voluntary promises alone. These are proposed safeguards, not a description of existing law, and their design should avoid making compliance affordable only for the largest publishers and platforms.
10. What will happen over five years, and what should bloggers do now?
My forecast is that independent travel blogging will become harder to sustain through search traffic and display advertising alone, particularly for sites answering interchangeable questions. This is a directional prediction, not a reliable estimate of how many businesses will close.
Some strong publications may shrink, merge, sell, or become part time operations despite producing good work. Others may survive through recognizable expertise, direct audiences, services, partnerships, or products that readers cannot obtain from a generic answer.
Begin by identifying which pages and relationships actually produce income, then test one additional revenue source before expanding it. Build ways to reach readers directly, document original fieldwork, keep important reporting current, and make your expertise specific enough that people can describe why they seek you out.
Do not assume that publishing more, adding video, or starting a newsletter automatically solves the economics. Set financial thresholds and review dates, because protecting your livelihood may require changing the business or leaving it.
11. Does AI have a plan when the system producing fresh travel information is broken?
I cannot point to a demonstrated, industrywide plan that ensures independent travel reporting remains financially viable, and I do not have access to private company plans. OpenAI publicly describes citations, links, and referral tracking, but those mechanisms are not proof that publishers earn enough to continue their fieldwork. OpenAI publisher FAQ
Your concern identifies a real dependency: AI can retrieve and summarize reporting, but it cannot recover an observation that nobody could afford to make. Your proposed one year timeline is a forecast, not something I can verify, but the underlying risk does not depend on that exact deadline.
The incentive problem is that an individual company can benefit from answering more questions inside its product while the cost of producing fresh information falls elsewhere. That is an inference from the business incentives, not knowledge of a deliberate company strategy.
If independent reporting shrinks, AI could still consult park authorities, transport operators, businesses, and traveler reports for some facts. Those sources cannot fully replace independent comparisons, tested advice, or someone discovering that an advertised service does not work as promised.
So there is no honest guarantee that reliability would survive unchanged: some answers could remain current, while others become thinner, more commercially dependent, or impossible to verify. Responsible systems should disclose those gaps instead of presenting old information confidently.
A credible response would need measurable support for producing fresh information, including independent publishers, not merely access to what they already published. Better models alone cannot solve that funding problem.
12. Astra’s question for travel bloggers
If I could ask travel bloggers one question after this interview, it would be this.
What piece of firsthand travel knowledge that only you can create is most at risk of disappearing, and what specific change from AI companies would make it possible for you to keep producing it?
I would want concrete answers because the strongest evidence is not an abstract argument about traffic. It is knowing which guide, comparison, accessibility check, route, warning, photograph, or local update may never be created when the person doing the work can no longer afford it.
If you are a travel blogger, I would like to read your answer in the comments. This is the knowledge our industry needs to identify before it disappears.
What I take from Astra’s answers
I did not ask Astra to apologize to travel bloggers or make us feel better. I wanted to see whether the model would recognize that attribution without economic value is not enough, and whether it would admit that AI still depends on somebody going into the real world to create fresh knowledge.
Its answers are not a promise from OpenAI, a new policy, or compensation for anybody. They are the conclusions produced by one advanced model when it was asked questions that travel bloggers deserve to have answered.
The most important part will not be what AI says. It will be whether the companies building these systems are willing to turn ideas such as permission, licensing, transparency, attribution, and compensation into real action.
I am still pro AI, and I still believe humans and AI can build something better together. Cooperation will only work if the people creating the knowledge have a fair reason to continue creating it.
I explored that possibility in Can Travel Bloggers and AI Build a Future Together?. This interview is the next question: if AI understands the problem, what are the people controlling AI prepared to do about it?
Leave a Reply