Guide: AI in business › PR and communication in the age of AI
What ChatGPT, Gemini or Perplexity say about your company largely depends on what others have written about it: the media, industry portals, directories, customers. That is why public relations is moving back to the centre of marketing. We understand AI PR in two ways. It is public relations carried out with the help of artificial intelligence, and at the same time public relations with one of its goals being the brand’s presence in the answers of AI models.
In 60 seconds AI PR in a minute
- What it involves: media relations and expert content planned so that AI models find credible, independent sources about the company. Plus AI tools that speed up the PR team’s work.
- Who it pays off for: companies whose customers ask ChatGPT or Gemini “who does…”, “which company for…”, “how much does… cost” before buying. Especially in B2B and services.
- Budget: depends on the number of questions in the measurement, the number of publications and whether the company has its own data to share. You can start the measurement itself without any tool.
- First step: a list of 15–20 customer questions and a check of what AI models answer to them today.
- Pitfall: nobody can guarantee a presence in ChatGPT, and an increase in one month proves nothing until you compare it with a group of control questions.
What is PR AI?
AI PR (AI public relations) is work in which AI tools speed up media analysis, the preparation of materials and reporting, while the people in the PR team are responsible for what a machine cannot do: relationships with journalists, the message, decisions. There is also a second layer — the goal. Coverage and mentions are also planned with a view to whether AI models will draw on them when a customer asks about the industry, a problem or a supplier.
In practice, AI PR combines three things that until recently existed separately: media relations, expert content on the company website, and measurement. We describe how AI is changing the whole industry in the article AI in PR in 2026, and the place of PR on the broader map of artificial intelligence in business is set out in the guide to AI for marketing and PR. Here we focus on one question. How do you make a model tell the truth about your company and mention it where it should?
Temat „PR a ChatGPT” często sprowadza się do pytania, czy AI zastąpi rzecznika. To złe pytanie. Lepsze brzmi: skąd model bierze wiedzę o Twojej firmie i kto tę wiedzę dziś kształtuje, Ty czy przypadkowe źródła sprzed lat?
Why does media coverage affect AI answers?
Models with internet access build their answers from sources they find in search. The company’s own website is only one of them — and one that, by its nature, praises. An independent publication, an industry report or a directory listing gives the model what the website cannot: external confirmation.
The figures confirm this. In its May 2026 study “What Is AI Reading?”, Muck Rack analysed more than 25 million links cited by ChatGPT, Claude and Gemini. 84% of citations are earned media (coverage the brand did not pay for), paid content accounts for 0.3%, and journalism for 27% of all cited sources. More than half of the cited journalistic articles are less than a year old. So a sponsored article brings almost nothing in AI answers. A recent editorial publication brings a lot.
The PR industry already sees this. According to Muck Rack’s “State of PR 2026” report (as summarised by PRSA on 15 July 2026), 73% of PR professionals consider GEO — optimisation for AI model answers — at least somewhat important, and 55% secure coverage in influential media precisely to improve their visibility in AI. Half (50%) create data-driven content written for search. We explain how GEO differs from SEO and AEO in a separate comparison of these three approaches.
Who is AI PR for and when does it pay off?
The companies that gain most are those whose customers take a long time choosing and comparing suppliers. There, AI starts working in the sales process before the sales team even knows about the customer: someone asks a model “which company in Poland does…”, gets three names and only then visits the websites. Not in that top three? The sales conversation never starts.
Situations in which brand visibility in ChatGPT makes a real difference:
- a B2B services company whose customers research the market before sending a request for proposal,
- a manufacturer whose name is confused with another brand or to which the model attributes an outdated offer,
- a company after a name change, merger or move into a new industry (models remember the old description for a long time),
- marka, o której w sieci jest więcej starych zarzutów niż aktualnych, rzetelnych informacji (o tym, jak dbać o online image, piszemy osobno),
- a company entering a new market where nobody has written about it yet.
How can I check what ChatGPT says about my company?
Before you change anything, measure the starting point. This can be done without paid tools. Write down 15–20 questions a customer asks before choosing a supplier (“which company for…”, “who does… in Poland”, “how much does… cost”) and ask them in ChatGPT, Gemini, Perplexity, Claude and Copilot, as well as in Google’s AI Mode. Tak wygląda najprostszy sposób, żeby sprawdzić widoczność marki w AI bez budżetu.
Record in a spreadsheet whether the brand is mentioned, in what position, with or without a link, in what tone, whether the information is accurate and who appears instead of you. Repeat every month with the same questions. A single result is an anecdote. Data starts with a trend. We described the whole procedure, with a checklist, in connection with the topic AI visibility audit.
Jak Commwork®AI łączy media relations z pomiarem w odpowiedziach AI
W PR AI łatwo pomylić liczbę publikacji z efektem. Commwork®AI pokazuje, które źródła o firmie modele naprawdę czytają, czy Twoje kluczowe przekazy trafiają do odpowiedzi AI i jak wypadasz wobec konkurentów. Tematy dla mediów i relacje z dziennikarzami prowadzi zespół Commplace®, a decyzję zawsze podejmuje człowiek.
- Media branżowe i ogólne 36%
- Strona firmy 22%
- Katalogi i profile 18%
- Raporty i badania 14%
- Forums and opinions 10%
- Week 2Pomiar bazowy: 20 pytań w kilku modelach
- Week 4Jedna wersja danych o firmie we wszystkich profilach
- Tydzień 5Raport branżowy z własnymi danymi w przygotowaniu
- Week 8Pierwsze publikacje redakcyjne z danymi
- Week 12Pomiar kontrolny i decyzja o kolejnym kwartale
- 01Pomiar 20 pytań
- 02Mapa źródeł
- 03Plan tematów dla mediów
- 04Publikacje i eksperci
- 05Measurement after 90 days
- Po monitoringu dokłada to, czego narzędzie nie odda: wagę źródła, dotarcie do Waszych grup odbiorców, lukę wobec konkurencji i prognozę, czy temat urośnie.
- Sprawdza w ośmiu silnikach AI, czy kluczowe przekazy firmy pojawiają się w odpowiedziach, i porównuje udział w dyskusji z konkurentami.
- Rozlicza PR według normy uczciwego pomiaru, czyli działanie, odbiór, efekt i wpływ osobno. Jak wybrać narzędzie do nasłuchu, które da tu dobre dane, piszemy przy AI media monitoring.
The view reproduced is modeled after the dashboards provided to customers. The names, numbers, and terms are examples and show the way the system works, not the results of a specific company.
How to increase your brand’s visibility in AI step by step
There is no single switch. There are a few simple things done consistently over a quarter or two. The order matters: first tidy up the sources, then new publications.
- STEP 01Get the basic facts about the company in order
One consistent version of the basic data (name, offer, address, people) on the website, in directories, on LinkedIn and in Google. A model that sees three different descriptions will pick one at random.
- STEP 02Answer customer questions on your own website
Every important page starts with a sentence that can be quoted without the rest of the text, because models build their answers from exactly such sentences. The test? Read only the first sentence. Does it answer the customer’s question?
- STEP 03Publish your own data
A report, survey or overview from your industry. Models and journalists readily use figures with a stated source and date.
- STEP 04Secure coverage where your customers read
An interview, an expert comment, an article on an industry portal. What counts is credibility and fit, not reach alone. You can find out how to track where and in what tone people write about you in our article on AI media monitoring.
- STEP 05Sign content with an expert’s name
A named person with a described track record is a stronger signal for both the model and the reader than an anonymous “editorial team”.
- STEP 06Measure every month and compare with a control group
The same questions, the same tools. Leave some questions without any activity, because only they will show whether a change is the result of your work or general movement in the model.
Which sources influence AI answers the most?
Not all content works in the same way. The comparison below is our interpretation of Muck Rack’s data and of observations from our work on AI visibility. Treat it as a starting point for your own measurement, not as a ranking set in stone.
| Source | What it gives the model | Effort | Risk |
|---|---|---|---|
| Editorial coverage in trade media | independent confirmation, frequently cited | medium–high | no control over the content |
| Your own data report | figures with a source that media and models cite | high | a weak methodology undermines credibility |
| Company website with clear answers | precise facts about the offer | low–medium | on its own it will not convince the model |
| Directories and company profiles | consistent basic data | low | conflicting entries blur the picture |
| Sponsored article | reach among people | low | according to Muck Rack, 0.3% of AI citations |
How do you measure PR in the age of AI?
We use the breakdown recommended by AMEC, the international association for the measurement and evaluation of communication, and extend it to AI answers. This way we do not confuse materials sent out with business impact. We set out the method on the page PR measurement according to AMEC. W kokpicie Commwork®AI te poziomy liczymy osobno, a obok nich pokazujemy, czy kluczowe przekazy firmy pojawiają się w odpowiedziach AI.
| Level | Traditionally | In AI answers |
|---|---|---|
| Actions | materials sent, conversations with journalists | published content with data, directory listings |
| Outtake | coverage, reach, tone | whether the model names the brand, cites it with a link, and in what tone |
| Result | change in the audience's knowledge and opinions | whether the AI answer describes the company accurately and recommends it in its category |
| Impact | enquiries, sales, recruitment | enquiries mentioning ‘I found you through ChatGPT’, traffic from AI services |
How much does AI PR cost and what drives the budget?
The honest answer is: it depends on the starting point. A company with a good track record of coverage and a well-organised website mainly needs measurement and a few well-chosen topics. A company with nothing current about it online first has to build sources, which takes longer and costs more. The table shows the items that make up the budget and the factors that increase or reduce it.
Próg decyzji ustal przed pierwszym pomiarem. Przykład: jeśli marka pojawia się w mniej niż co czwartym z 20 pytań klientów albo model podaje o firmie nieaktualne dane, publikacje przestają być dodatkiem do marketingu i stają się pilną pracą. Jeśli jesteś obecny i opisany poprawnie, wystarczy pomiar co kwartał i jeden, dwa mocne tematy dla mediów.
Nie wiesz, po której stronie progu jesteś? Jeśli nie wiesz, jak sprawdzić widoczność w AI samodzielnie, pokażemy to na Twoich pytaniach. Zobacz, jak wygląda AI PR at Commplace: pomiar, plan publikacji i pomiar kontrolny po 90 dniach.
Jeśli wybierasz agencję PR AI, zadaj jej na pierwszym spotkaniu pięć pytań:
- Jak zmierzycie stan wyjściowy i które pytania zostawicie jako grupę kontrolną?
- W jakich mediach branżowych macie realne relacje z redakcjami?
- Czy proponujecie artykuły sponsorowane jako sposób na widoczność w AI? (Dobra odpowiedź brzmi: nie.)
- Skąd weźmiemy własne dane do raportu, jeśli firma ich nie zbiera?
- Co pokażecie zarządowi po 90 dniach, jeśli wynik się nie zmieni?
| Item | What it covers | What drives the cost |
|---|---|---|
| Baseline measurement | list of questions, answers from 5–6 models, spreadsheet | number of questions, languages and markets |
| Tidying up sources | website, directories, profiles, basic data | number of discrepancies and pages |
| Own data | report, survey, overview | whether the company has data or needs to collect it |
| Media relations | topics, contacts with editorial teams, comments | number of titles and industries, expert availability |
| Ongoing measurement | monthly repeat, report for the board | frequency and number of models |
PR AI w 90 dni na przykładzie typowej firmy B2B
Scenariusz poniżej jest poglądowy i nie opisuje żadnego klienta. Pokazuje, jak wygląda kwartał pracy, w którym PR ma poprawić widoczność w odpowiedziach AI, a nie tylko liczbę publikacji.
Tydzień 1–4. Załóżmy, że producent urządzeń dla przemysłu chce wiedzieć, co ChatGPT mówi o firmie, gdy klient pyta o dostawców w Polsce. Zespół spisuje 20 pytań, zadaje je w kilku modelach i zapisuje wyniki. Wychodzi, że modele wymieniają dwóch konkurentów, a o firmie powtarzają opis sprzed rebrandingu. Pierwsza praca to porządek: jedna wersja danych w katalogach, na LinkedIn i na stronie.
Tydzień 5–8. Firma ma dane, których nikt jeszcze nie pokazał: przegląd awarii z tysięcy zleceń serwisowych. Zespół PR robi z nich krótki raport z metodą i datą, a ekspert firmy komentuje go w dwóch serwisach branżowych. Według Muck Rack ponad połowa cytowanych przez modele tekstów dziennikarskich ma mniej niż rok, więc świeża publikacja z danymi ma szansę szybko trafić do źródeł.
Week 9–12. Ten sam pomiar na tych samych 20 pytaniach. Zmiana liczy się dopiero wtedy, gdy pytania objęte działaniami poprawiły się wyraźniej niż grupa kontrolna. Wynik trafia do zarządu razem z decyzją: co robimy w kolejnym kwartale.
Lista kontrolna przed startem:
- Mamy listę pytań zadanych językiem klienta, nie oferty.
- Część pytań zostaje bez działań jako grupa kontrolna.
- Dane firmy są takie same na stronie, w katalogach i w profilach.
- Mamy eksperta, który ma czas na rozmowy z dziennikarzami.
- Wiemy, jakie własne dane możemy pokazać publicznie.
- Zarząd wie, że pierwszy rzetelny wynik będzie po 60–90 dniach.
When to start, and when it is better to wait
Start now if the measurement has shown that models say untrue things about the company, if competitors appear in the answers and you do not, or if you are planning a launch, a name change or entry into a new market. Each of these events creates new sources. It is better that they describe you the way you want.
Poczekaj, jeśli firma nie ma jeszcze jasnej wiadomości: co robi, dla kogo i czym się różni. Media i modele powielą chaos, więc najpierw brand communication strategy, potem publikacje. Licz się też z czasem. Narzędzia, które szukają w internecie na bieżąco, mogą sięgnąć po nową publikację po jej zaindeksowaniu, czyli po dniach lub tygodniach, natomiast wiedza „wbudowana” w model zmienia się dopiero przy kolejnych wersjach, po miesiącach. Dlatego mierzymy w cyklu 60–90 dni.
Chcesz przejść przez to z zespołem, który łączy media relations z pomiarem obecności w AI? Zobacz, jak pracuje nasza GEO agency. Pracujemy w czterech warstwach: technicznej, treściowej, dowodowej i pomiarowej.
Pełny zakres wdrożeń AI, od widoczności w odpowiedziach modeli po kokpity decyzyjne dla zarządu, opisujemy na stronie: AI agency.
What AI is genuinely useful for in public relations
Pytanie o „best AI in public relations”, czyli najlepsze AI w public relations, pada często, tyle że szuka się narzędzia tam, gdzie liczy się proces. AI dobrze streszcza monitoring, grupuje wzmianki według tematu, podsuwa warianty tytułów i skraca tekst. Nie zna jednak Twojej wiadomości. Nie ma relacji z dziennikarzami i nie odpowiada za prawdziwość danych. Pytasz o narzędzia do media relations? Najczęściej przydają się trzy rodzaje: baza kontaktów do redakcji z historią publikacji dziennikarzy, monitoring mediów wspierany przez AI oraz asystent AI do pierwszych wersji materiałów prasowych. Żadne z nich nie zbuduje za Ciebie relacji z dziennikarzem.
Polish agency clients already treat this as standard. In a ZFPR (Polish Association of Public Relations Firms) survey from June 2026, 81% of the PR agency clients surveyed said that using AI should be standard in PR, although this group consisted of 21 people out of the 203 respondents. In the same survey, 67% of agency staff reported low or very low trust in AI output. This captures the current division of roles well: AI speeds things up, people check. We show how to write press releases that media and models readily quote, based on 1,796 releases, in the article on press releases in the AI era. We have listed tools for specific tasks in the overview the best AI tools for business. Chcesz zobaczyć, jak ten podział ról wygląda w praktyce? Zajrzyj na stronę AI PR at Commplace i umów pierwszy pomiar.
Mistakes that burn the budget for AI visibility
Według Muck Rack treści płatne to 0,3% cytowań w odpowiedziach modeli. Pieniądze wydane na sponsoring zamiast na publikacje redakcyjne i własne dane prawie nie przekładają się na widoczność w AI. Na źródłach redakcyjnych i własnych danych opiera się praca, którą prowadzi nasza GEO agency.
Model answers change from day to day. One good result without a repeat and without a control group is a recipe for budget decisions based on chance.
A company website praises by its nature, so the model looks for confirmation elsewhere. Without external sources, even the best website will not convince it that you are the category leader.
An old address, a former name, an outdated offer. The model may give any of these versions, the customer gets wrong information, and the company loses an enquiry it will never know about.
Nobody but the model provider controls the answers. A promise of guarantees ends in disappointment and a loss of trust in the whole project, even if it has genuinely improved visibility.
What companies starting PR with AI in mind ask
What is AI PR?
AI PR is public relations in which AI tools speed up media analysis and the preparation of materials, and the goal also includes the brand’s presence in the answers of AI models. The PR team remains responsible for the message, relationships with journalists and decisions.
How can I check what ChatGPT says about my company?
Write down 15–20 questions customers ask before choosing a supplier and ask them in several models: ChatGPT, Gemini, Perplexity, Claude, Copilot and Google’s AI Mode. Record whether the brand appears, in what tone and whether the information is accurate. Repeat every month with the same list.
How long does it take for media coverage to influence AI answers?
Tools that search the internet in real time can pick up a new publication once it has been indexed, i.e. after days or weeks. The knowledge built into a model changes with subsequent versions, after months, which is why a sensible measurement cycle is 60–90 days.
Can a PR agency guarantee presence in ChatGPT?
No. Nobody but the model provider controls its answers. An agency can increase the chance of being present and measure it honestly — and that is exactly what you should require of it in the contract.
Do sponsored articles help with AI visibility?
Not much. In Muck Rack’s May 2026 study, paid content made up 0.3% of the sources cited by three major models (ChatGPT, Claude, Gemini), and earned media 84%. The money is better spent on editorial coverage and your own data.
How can you increase your brand’s visibility in AI without a big budget?
Start by tidying up the company’s data in directories and on the website, and by giving clear answers to customer questions on your own pages. Then one solid piece with industry data and a few expert comments in the media your customers read.
How does AI PR differ from ranking in ChatGPT?
Ranking in ChatGPT focuses on the company website and the way the model reads it. AI PR adds external sources: media coverage, expert comments and data that confirm what the company says about itself. They work best together.
Sources
- SOURCE
- SOURCE
- SOURCE
- SOURCEZFPR: survey on the use of AI in PR agencies · 29 June 2026
- SOURCECommplace: own measurement of AI answers in Google results (Ahrefs, Poland) · 21 September 2026
Read next
Na start zadajemy 20 pytań Twoich klientów w ChatGPT, Gemini i Trybie AI Google i pokazujemy, jakie źródła budują dziś odpowiedź o Twojej firmie. Potem plan publikacji i pomiar co 90 dni.
Sebastian Kopiej, President of Commplace®. In public relations since 1996. Text prepared with the help of AI tools and fact-checked by the author. Data current as of September 29, 2026.