Even people with mild symptoms of the coronavirus develop antibodies, according to a study that raises hopes that survivors do get immunity.
Researchers analysed the blood of 160 medics who all had a mild form of Covid-19 and had not needed hospital care.
They found 99.4 per cent of the group had antibodies to the virus 13 days after their symptoms, which signals they have some form of protection against it.
The antibodies had the ability to 'neutralise' - or kill - the virus in tests, and appeared to grow stronger in the weeks after infection.
The study did not look at blood samples any later than 40 days after symptoms, so it was not established how long antibodies last for.
Antibodies are a sign a person was infected by the coronavirus. But whether or not this protects a survivor from being re-infected has yet to be proven.
Several countries, including the UK, have hung their hopes on tests that identify coronavirus antibodies to decide who is immune and can go back to work with so-called 'immunity passports'.
But the unsolved question regarding immunity has halted this idea and become one of the most contested issues during the pandemic.
The French study comes after previous research revealed that mild Covid-19 patients have such low levels of antibodies they are barely detectable, which could mean they are often missed by antibody tests, or that person has weak immunity.
The latest study was conducted at the Pasteur Institute, in Paris. The findings, which have yet to be peer-reviewed, have been published on MedRxiv.
Blood samples were taken from more than 160 staff at two hospitals in Strasbourg, eastern France, who became infected with Covid-19 early in March.
The cluster emerged after a 'super-propagation' event - linked to a five-day prayer and fasting gathering of the evangelical Christian Open Door church in Mulhouse, 70 miles (114km) from Strasbourg.
Participants - doctors, nurses, dentists and other medics - were 32 years old on average, and anyone who was severely sick with the disease had been removed from the study.
The researchers wanted to focus on patients with mild disease because there is little known about their immunity even though they make up to 80 per cent of cases, according to the World Health Organisation.
Two separate antibody tests were used to look for antibodies. The first test was a commercially available one, which generally are less accurate.
The second test had been developed by the institute to identify antibodies, and the ability of those antibodies to neutralise the virus.
The first test, a rapid immunodiagnostic test, detected antibodies in 153 (95.6 per cent) of the samples.
The more robust test found antibodies in 159 (99.4 per cent) volunteers. It failed to detect antibodies in a 58-year-old man.
When looking at the antibodies' ability to neutralise the virus, they appeared to grow stronger over time.
Neutralizing antibodies - which bind to the virus and inactivate it, rather than 'tag' it for other immune cells - were found in 79 per cent of samples 13-20 days after symptoms started. This rose to 98 per cent after 28-41 days.
'It is a fair assumption that the majority of individuals with mild Covid-19 generate neutralising antibodies within a month after onset of symptoms,' Olivier Schwartz, one of the leading researchers, told the French newspaper Le Monde.
'The neutralising activity is present much later than the appearance of antibodies and this is encouraging.
'Although not yet demonstrated, several lines of evidence suggest that the presence of neutralising antibodies may be associated with protective immunity for Sars-CoV-2 infection.'
Survivors of SARS, a coronavirus related to this one that died out in early 2000, have been found to have antibodies that are protective for two years.
The evidence that mild Covid-19 patients elicit the production of antibodies has not always been positive. Some research has shown only patients with really severe disease develop antibodies, which may mean those who had mild forms are still vulnerable to infection again.
A study of a similar size at Rockefeller University in New York City, published this week, also found that every patient's immune system seemed to be capable of generating the types of antibodies that neutralize the virus - but sometimes in very weak amounts.
For the study, published on pre-peer review site bioRxiv.org, the team looked at 149 people who donated blood plasma - which would contain antibodies.
Researchers then mixed the plasma with the spike protein of SARS-CoV-2 - which coats the virus - and measured if or how well the virus would infect human cells in a petri dish.
Most samples, which were collected 39 days after symptoms on average, did not do very well at neutralizing the virus.
In fact, the neutralizing effect was undetectable in 33 per cent of donors. The investigators say this may be because their immune systems cleared the infection before antibodies could be produced.
Antibodies are only one part of the picture - there are other parts of the immune system which are triggered by the virus, including T cells.
Health officials in several countries, including the UK, have hung their hopes on tests that identify coronavirus antibodies to decide who is immune and can go back to work.
But getting the ball rolling on antibody testing has been a fiasco, not only because the science on immunity is full of holes, but because there have been obstacles in finding a test with enough reliability.
Most antibody tests are fraught with inaccuracies, giving an incorrect result, which can be dangerous.
Yesterday the Centers for Disease Control found antibody tests for Covid-19 may be wrong up to half of the time, and warned testing is not accurate enough for it to be used for any policy-making decisions.
NHS workers in the UK were confirmed to be getting antibody testing this week after a deal to buy 10million tests from manufacturers Roche and Abbott.
These were planned to be rolled out to the public soon after - but there are concerns this will not go ahead.
Health officials have banned people from buying the same tests online because they fear the results are too unreliable.
The ban affects Superdrug and Lloyds, which were just two companies offering the tests to people for a fee of £69 or more using the same equipment that are now being used for NHS staff.
The equipment is being used wrongly, officials say, because it takes blood from a finger prick sample - not someone's veins, as been approved by the Government.
There has been no rapid home testing kit - which someone reads results on like a pregnancy stick - approved for use in the UK by health chiefs, despite promises back in March.